Snap DSA Report Risk Assessment Results and Mitigations August 2024 CONFIDENTIAL Contents Foreword.................................................................................................................................................12 What is New?.......................................................................................................................................... 14 Previous reports.............................................................................................................................................. 14 What is new in this Report?.......................................................................................................................... 14 Section 1 - Introduction........................................................................................................................... 14 Section 2 - DSA Risk Assessment Scope.......................................................................................... 14 Section 3 - DSA Risk Assessment Methodology............................................................................. 15 Section 4 - DSA Risk Assessment Results.........................................................................................16 Section 5 - Specific Mitigations............................................................................................................ 17 Section 6 - Ongoing Risk Detection \& Management..................................................................... 19 Annex - Explainer Series.......................................................................................................................20 Conclusion................................................................................................................................................20 1\. Introduction........................................................................................................................................20 1.1 Snapchat 1.01............................................................................................................................................... 21 1.2 Snapchat Community..............................................................................................................................23 2\. DSA Risk Assessment Scope.......................................................................................................... 25 2.1 Approach.................................................................................................................................................... 25 2.1.1 Scope in our 2023 Report............................................................................................................ 26 2.1.2 Scope for this Report....................................................................................................................26 2.2 Spotlight.................................................................................................................................................... 26 2.3 Discover.....................................................................................................................................................27 2.4 Public Profiles.......................................................................................................................................... 28 2.5 Snap Map.................................................................................................................................................. 29 2.6 Lenses........................................................................................................................................................ 34 2.7 Advertising................................................................................................................................................ 35 3\. DSA Risk Assessment Methodology.............................................................................................. 37 3.1 Identification of Risks...............................................................................................................................37 3.2 Likelihood Analysis................................................................................................................................. 38 3.3 Severity Analysis..................................................................................................................................... 39 3.4 Overall Potential Risk Prioritization Assessment............................................................................ 39 3.5 Snap's Mitigations................................................................................................................................... 40 3.6 Conclusions............................................................................................................................................... 41 3.7 Supporting Documentation...................................................................................................................42 4\. DSA Risk Assessment Results........................................................................................................ 43 4.1 Category 1 - Dissemination of content that is illegal or violates our terms and conditions.. 43 4.1.1 Dissemination of child sexual abuse material......................................................................... 45 2 CONFIDENTIAL Likelihood...........................................................................................................................................45 Severity...............................................................................................................................................46 Overall potential risk prioritization.............................................................................................. 46 Snap's Mitigations............................................................................................................................ 46 Conclusion......................................................................................................................................... 49 4.1.2 Dissemination of illegal hate speech....................................................................................... 50 Likelihood.......................................................................................................................................... 50 Severity...............................................................................................................................................50 Overall potential risk prioritization............................................................................................... 51 Snap's Mitigations............................................................................................................................. 51 Conclusion......................................................................................................................................... 53 4.1.3 Dissemination of information related to the sale of prohibited products or services. 54 Likelihood...........................................................................................................................................54 Severity...............................................................................................................................................54 Overall potential risk prioritization.............................................................................................. 55 Snap's Mitigations............................................................................................................................55 Conclusion.........................................................................................................................................58 4.1.4 Dissemination of terrorist content............................................................................................. 59 Likelihood.......................................................................................................................................... 59 Severity...............................................................................................................................................60 Overall potential risk prioritization.............................................................................................. 60 Snap's Mitigations............................................................................................................................60 Conclusion......................................................................................................................................... 63 4.1.5 Dissemination of content that infringes on intellectual property rights.......................... 64 Likelihood...........................................................................................................................................64 Severity...............................................................................................................................................65 Overall potential risk prioritization.............................................................................................. 65 Snap's Mitigations............................................................................................................................65 Conclusion......................................................................................................................................... 67 4.1.6 Dissemination of adult sexual content..................................................................................... 68 Likelihood.......................................................................................................................................... 68 Severity...............................................................................................................................................69 Overall potential risk prioritization.............................................................................................. 69 Snap's Mitigations............................................................................................................................69 Conclusion......................................................................................................................................... 72 4.1.7 Dissemination of content regarding harassment \& bullying............................................... 72 Severity............................................................................................................................................... 74 Overall potential risk prioritization...............................................................................................74 Snap’s Mitigations............................................................................................................................ 74 3 CONFIDENTIAL Conclusion..........................................................................................................................................77 4.1.8 Dissemination of content that glorifies self-harm, including suicide................................ 77 Likelihood...........................................................................................................................................78 Severity............................................................................................................................................... 78 Overall potential risk prioritization...............................................................................................79 Snap's Mitigations............................................................................................................................ 79 Conclusion......................................................................................................................................... 82 4.1.9 Dissemination of content relating to violent or dangerous behavior.............................. 82 Likelihood...........................................................................................................................................83 Severity...............................................................................................................................................84 Overall potential risk prioritization...............................................................................................84 Snap's Mitigations............................................................................................................................ 84 Conclusion......................................................................................................................................... 87 4.1.10 Dissemination of harmful false information........................................................................... 88 Likelihood...........................................................................................................................................88 Severity...............................................................................................................................................88 Overall potential risk prioritization.............................................................................................. 88 Snap's Mitigations............................................................................................................................89 Conclusion......................................................................................................................................... 93 4.1.11 Dissemination of fraud and spam..............................................................................................93 Likelihood...........................................................................................................................................94 Severity...............................................................................................................................................95 Overall potential risk prioritization.............................................................................................. 95 Snap's Mitigations............................................................................................................................95 Conclusion.........................................................................................................................................98 4.1.12 Dissemination of information related to other illegal activities........................................ 98 Likelihood.......................................................................................................................................... 98 Severity...............................................................................................................................................99 Overall potential risk prioritization.............................................................................................. 99 Snap's Mitigations............................................................................................................................99 Conclusion....................................................................................................................................... 102 4.2 Category 2: Negative Effects on Fundamental EU Rights..........................................................102 4.2.1 Right to human dignity................................................................................................................104 Likelihood......................................................................................................................................... 104 Severity............................................................................................................................................. 105 Overall potential risk prioritization.............................................................................................106 Snap's Mitigations.......................................................................................................................... 106 Conclusion....................................................................................................................................... 109 4.2.2 Right to freedom of expression and assembly................................................................... 109 4 CONFIDENTIAL Likelihood.........................................................................................................................................109 Severity.............................................................................................................................................. 110 Overall potential risk prioritization..............................................................................................110 Snap's Mitigations........................................................................................................................... 110 Conclusion.........................................................................................................................................113 4.2.3 Right to private life.......................................................................................................................114 Likelihood.......................................................................................................................................... 114 Overall potential risk prioritization.............................................................................................. 114 Snap's Mitigations........................................................................................................................... 115 Conclusion.........................................................................................................................................118 4.2.4 Right to data protection..............................................................................................................118 Likelihood.......................................................................................................................................... 118 Severity.............................................................................................................................................. 119 Overall potential risk prioritization.............................................................................................. 119 Snap's Mitigations........................................................................................................................... 119 Conclusion........................................................................................................................................123 4.2.5 Right to non-discrimination and freedom of religion.........................................................123 Likelihood......................................................................................................................................... 124 Severity............................................................................................................................................. 124 Overall potential risk prioritization.............................................................................................125 Snap's Mitigations.......................................................................................................................... 125 Conclusion.........................................................................................................................................131 4.2.6 Children’s Rights...........................................................................................................................131 Likelihood......................................................................................................................................... 132 Severity............................................................................................................................................. 132 Overall potential risk......................................................................................................................133 Snap's Mitigations...........................................................................................................................133 Conclusion........................................................................................................................................136 4.2.7 Right to consumer protection...................................................................................................136 Likelihood.................................................................................................................137 Severity..............................................................................................................................................137 Severity.................................................................................................................... 137 Overall potential risk prioritization............................................................................................. 137 Snap's Mitigations...........................................................................................................................138 Conclusion.........................................................................................................................................141 4.2.8 Right to Property.......................................................................................................................... 141 4.3 Category 3: Negative effect on public security.............................................................................. 141 4.3.1 Negative Effect on Democratic and Electoral Processes.................................................. 142 Likelihood......................................................................................................................................... 142 5 CONFIDENTIAL Severity..............................................................................................................................................143 Overall potential risk prioritization............................................................................................. 144 Snap's Mitigations...........................................................................................................................144 Conclusion....................................................................................................................................... 156 4.3.2 Negative Effect on Civil Discourse.........................................................................................156 Likelihood......................................................................................................................................... 157 Severity..............................................................................................................................................157 Overall potential risk prioritization............................................................................................. 157 Snap's Mitigations...........................................................................................................................157 Conclusion.........................................................................................................................................161 4.3.3.Negative Effect on Public Security......................................................................................... 162 Likelihood......................................................................................................................................... 162 Severity............................................................................................................................................. 163 Overall potential risk prioritization............................................................................................. 163 Snap's Mitigations...........................................................................................................................163 Conclusion....................................................................................................................................... 168 4.4 Category 4: Negative Effects on Public Health............................................................................. 168 4.4.1 Negative Effects on Public Health........................................................................................... 169 Likelihood......................................................................................................................................... 169 Severity............................................................................................................................................. 170 Overall potential risk prioritization............................................................................................. 170 Snap's Mitigations...........................................................................................................................170 Conclusion........................................................................................................................................173 4.4.2 Negative Effects on gender-based violence........................................................................174 Likelihood......................................................................................................................................... 174 Severity..............................................................................................................................................174 Overall potential risk prioritization............................................................................................. 174 Snap's Mitigations...........................................................................................................................175 Conclusion........................................................................................................................................ 177 4.4.3 Negative Effects on Children................................................................................................... 178 Likelihood......................................................................................................................................... 178 Severity..............................................................................................................................................179 Overall potential risk prioritization............................................................................................. 179 Snap's Mitigations...........................................................................................................................179 Conclusion....................................................................................................................................... 182 4.4.4 Serious Negative Consequences on physical and mental well-being......................... 183 Likelihood......................................................................................................................................... 183 Severity............................................................................................................................................. 186 Overall potential risk prioritization.............................................................................................186 6 CONFIDENTIAL Snap’s Mitigations...........................................................................................................................187 Conclusion.........................................................................................................................................191 5\. Specific Mitigations........................................................................................................................ 192 5.1 Snapchat Design and Function........................................................................................................... 193 5.1.1 Adaptations and Mitigations....................................................................................................... 193 Spotlight and Discover..................................................................................................................194 Public Profile....................................................................................................................................195 Snap Map..........................................................................................................................................197 Lenses............................................................................................................................................... 198 Advertising....................................................................................................................................... 199 5.1.2 Integrations with other mitigations..........................................................................................199 5.1.3 Online Interface Design Process............................................................................................ 200 5.1.4 Online Design Principles...........................................................................................................200 5.1.5 Conclusion....................................................................................................................................200 5.2 Terms....................................................................................................................................................... 200 5.2.1 Introduction.................................................................................................................................. 200 5.2.2 Terms and Conditions................................................................................................................201 Terms of Service............................................................................................................................. 201 Community Guidelines.................................................................................................................204 Privacy Policy..................................................................................................................................207 Advertising Policies...................................................................................................................... 208 5.2.3 Product Specific Terms.............................................................................................................209 Spotlight...........................................................................................................................................209 Discover...........................................................................................................................................209 Lenses............................................................................................................................................... 210 5.2.4 Other Aspects..............................................................................................................................210 Oversight and Administration..................................................................................................... 210 Accessing Terms and Conditions...............................................................................................210 Languages........................................................................................................................................212 Readability........................................................................................................................................212 5.2.5 Conclusion....................................................................................................................................213 5.3 Transparency.......................................................................................................................................... 213 5.3.1 Information we provide on our website................................................................................. 214 Privacy, Safety, and Policy Hub...................................................................................................214 Policy Center................................................................................................................................... 215 Privacy Center.................................................................................................................................215 Safety Center................................................................................................................................... 217 Parents.............................................................................................................................................. 219 Transparency Center....................................................................................................................220 7 CONFIDENTIAL News Page....................................................................................................................................... 221 5.3.2 Information provided in app stores.......................................................................................222 5.3.3 Information we provide in our application...........................................................................223 Onboarding process.....................................................................................................................223 Just-in-time notifications..............................................................................................................225 Thematic awareness and notices............................................................................................. 226 5.3.4 Languages................................................................................................................................... 229 5.3.5 Conclusion...................................................................................................................................229 5.4 Content Moderation.............................................................................................................................229 5.4.1 Approach.......................................................................................................................................229 Snapchat Design and Function................................................................................................. 230 Community Guidelines and Terms of Service........................................................................230 Content Moderation..................................................................................................................... 230 Enforcement.....................................................................................................................................231 5.4.2 Content Moderation................................................................................................................... 231 Proactive Moderation (Content Reviews).................................................................................231 5.4.3 Conclusion.................................................................................................................................... 231 5.5 Enforcement.......................................................................................................................................... 232 5.5.1 Introduction...................................................................................................................................232 5.5.2 Protections against Misuse (Art. 23)..................................................................................... 232 5.5.3 Transparency for Measures of Protection........................................................................... 232 5.5.4 Notification of Criminal Offenses (Art. 18)............................................................................ 233 Proactive referrals to law enforcement and governmental agencies............................. 233 Law enforcement takedown requests (Articles 9 and 10)...................................................233 5.5.5 Complaint Handling System (Art. 20)................................................................................... 233 Statement of Reasons (Art. 17)....................................................................................................234 Notice to Reporter.........................................................................................................................234 5.5.6 Effectiveness of Enforcement.................................................................................................235 5.5.7 Conclusion................................................................................................................................... 236 5.6 Algorithmic Systems............................................................................................................................ 236 5.6.1 Introduction...................................................................................................................................236 5.6.2 Content Recommendation Systems......................................................................................237 How do our Content Recommender Systems work?...........................................................237 Benefits............................................................................................................................................ 238 5.6.3 Oversight and Administration................................................................................................. 238 Algorithmic System Review........................................................................................................ 238 5.6.4 Adaption and Testing................................................................................................................ 239 Summary..........................................................................................................................................239 Illegal or violating content...........................................................................................................239 8 CONFIDENTIAL Lack of user understanding........................................................................................................ 239 Intrusive personalized recommendations.............................................................................. 240 Discrimination..................................................................................................................................241 Rapid spread of illegal or false content \& crisis exposure.................................................. 241 Filter bubbles...................................................................................................................................241 Erroneously excluding content.................................................................................................. 242 Viewers could be watching but not enjoying content.........................................................242 5.6.5 Change Management............................................................................................................... 243 5.6.6 Monitoring and Quality Assurance........................................................................................ 243 Performance Monitoring.............................................................................................................. 243 Quality Assurance......................................................................................................................... 243 5.6.7 Conclusion....................................................................................................................................243 5.7 Advertising Systems.............................................................................................................................244 5.7.1 Introduction....................................................................................................................................244 5.7.2 How do our Advertising Systems Work?.............................................................................. 244 5.7.3 Benefits..........................................................................................................................................246 5.7.4 Adaptation and Testing..............................................................................................................247 Reasonable and Proportionate Targeting............................................................................... 247 Advertising Policies.......................................................................................................................248 Advertising Review....................................................................................................................... 248 Advertising Reporting...................................................................................................................249 Ad Markers......................................................................................................................................250 Transparency and Control...........................................................................................................252 Ads Gallery......................................................................................................................................253 Freedom of Expression................................................................................................................255 5.7.5 Conclusion....................................................................................................................................255 5.8 Protection of Minors............................................................................................................................ 255 5.8.1 Introduction...................................................................................................................................255 5.8.2 Overview and Approach..........................................................................................................256 Age Appropriate Design Code.................................................................................................. 256 Privacy, Safety, and Security of Minors on Snapchat........................................................... 257 Advertisements for Minors..........................................................................................................258 Identifying Minors..........................................................................................................................258 Ongoing evaluation...................................................................................................................... 264 Transparency to Minors............................................................................................................... 266 5.8.3 App Store Level Safeguards................................................................................................... 267 5.8.4 Device-Level Safeguards......................................................................................................... 267 5.8.5 Platform-Level Safeguards......................................................................................................269 5.8.6 Product-Level Safeguards....................................................................................................... 270 9 CONFIDENTIAL Public Content................................................................................................................................270 Viewing Public Content............................................................................................................... 270 Spotlight...........................................................................................................................................270 Map................................................................................................................................................... 270 Advertisements............................................................................................................................... 271 Reporting and Blocking............................................................................................271 Friending........................................................................................................................................... 271 Family Center / Parent Tools.......................................................................................................273 5.8.7 Conclusion....................................................................................................................................276 5.9 Content Authenticity............................................................................................................................ 277 5.9.1 Introduction................................................................................................................................... 277 5.9.2 Risk Assessment Results..........................................................................................................277 5.9.3 Mitigations....................................................................................................................................278 Guidelines, policies, and practices........................................................................................... 278 User Guidance................................................................................................................................ 281 Enforcement................................................................................................................................... 284 Partnerships.................................................................................................................................... 284 5.9.4 Conclusion................................................................................................................................... 285 5.10 Trusted Flaggers................................................................................................................................. 286 5.10.1 Trusted Flagger Program......................................................................................................... 286 5.10.2 Conclusion..................................................................................................................................287 5.11 Dispute Settlement Bodies................................................................................................................ 287 5.11.1 Overview and Approach............................................................................................................287 5.11.2 Enquires....................................................................................................................................... 288 5.11.3 Conclusion...................................................................................................................................288 5.12 Codes and Crisis Protocols.............................................................................................................. 288 5.12.1 Cooperation.................................................................................................................................288 5.12.2 Codes of Practice..................................................................................................................... 290 EU hate speech Code..................................................................................................................290 FSM Code of Conduct................................................................................................................. 290 EU disinformation code............................................................................................................... 290 Article 28 Guidance / EU AAD Code.........................................................................................291 5.12.3 Crisis Protocols..........................................................................................................................292 6\. Ongoing Risk Detection and Management................................................................................ 292 6.1 Platform Principles-based Framework............................................................................................. 293 6.2 DSA Compliance Team and Cross-Functional Working Groups..............................................293 6.2.1 Introduction...................................................................................................................................293 6.2.2 DSA Independent Compliance Function.............................................................................293 6.2.3 Independent Compliance Function Leads..........................................................................294 10 CONFIDENTIAL 6.2.4 Compliance Officer Qualifications.........................................................................................294 6.2.5 Operation of the Independent Compliance Function...................................................... 294 Responsibilities of the Independent Compliance Function............................................... 294 Oversight and Monitoring of Snap’s DSA Compliance....................................................... 295 DSA Management Body.............................................................................................................. 295 6.2.6 DSA Cross-Functional Governance Team...........................................................................295 6.2.7 Points of Contact........................................................................................................................ 296 Designation, Publication, and Change Management.......................................................... 296 Point of Contact for the Authorities.......................................................................................... 296 Point of Contact for Users........................................................................................................... 297 Legal Representative....................................................................................................................297 6.2.8 DSA Supervisory Fee................................................................................................................297 6.3 Privacy and Safety by Design........................................................................................................... 298 6.3.1 DSA Risk Management..............................................................................................................298 6.3.2 Privacy and Safety by Design review process...................................................................299 6.3.3 Holistic Digital Risk Management..........................................................................................299 6.3.4 DSA Critical Impact Check...................................................................................................... 300 6.4 Prevalence Testing...............................................................................................................................300 6.4.1 Overall............................................................................................................................................300 6.4.2 Example mitigations................................................................................................................... 301 6.4.3 Conclusion.................................................................................................................................... 301 6.5 External Request Monitoring and Review.......................................................................................301 6.6 Digital Well-Being Index (DWBI) Initiative.......................................................................................302 6.6.1 Introduction...................................................................................................................................302 6.6.2 Snap’s Digital Well-Being Index - Year Three.....................................................................302 6.7 Snap Advisory Groups.........................................................................................................................303 6.7.1 Introduction....................................................................................................................................303 6.7.2 Updates......................................................................................................................................... 303 Safety Advisory Board..................................................................................................................303 Snap Council for Digital Well-Being..........................................................................................304 6.8 Audit.........................................................................................................................................................306 7\. Conclusion....................................................................................................................................... 307 8\. Final Words...................................................................................................................................... 310 Annex 1 - Community Guidelines: Explainer Series.........................................................................311 Sexual Content............................................................................................................................................... 311 Harassment \& Bullying................................................................................................................................. 311 Threats, Violence \& Harm............................................................................................................................311 Harmful False or Deceptive Information..................................................................................................311 Illegal or Regulated Activities..................................................................................................................... 311 11 CONFIDENTIAL Hateful Content, Terrorism, and Violent Extremism............................................................................. 311 Severe Harm................................................................................................................................................... 311 Snapchat Content Moderation, Enforcement, and Appeals...............................................................311 Foreword This Risk Assessment Results and Mitigations Report (Report) has been prepared to comply withSnap’s obligations under Article 42.4.(a), (b) and (e) of Regulation (EU) 2022/2065 on a SingleMarket For Digital Services and amending Directive 2000/31/EC (the “Digital Services Act” or“DSA”). This Report is divided into eight sections: (1) Introduction; (2) DSA Risk Assessment Scope; (3)DSA Risk Assessment Methodology; (4) DSA Risk Assessment Results; (5) Specific Mitigations; (6)Ongoing Risk Detection and Management; (7) Conclusion; and (8) Final Words. Since our 2023 Report, we have made a number of significant updates and these aresummarized in a “What is New?” section that we have added to this Report. At a high level, theseupdates reflect three many themes that we have observed during the last year: ● The popularization of Gen AI - There has been intense interest and concern surroundingthe ways in which advancements and access to generative AI technologies are impactingonline platforms. Although our Gen AI creation tools are platforms generally not onlineplatforms and are outside the scope of this Report, in the light of this focus, we havenevertheless provided details of our mitigation measures for both creation anddissemination of Gen AI content. For more information, see Section 5.9 ContentAuthenticity in this Report. ● Elections and democratic process - 2024 has been described as a Global Elections SuperCycle, with a reported 73 different national elections. For the EU, the EuropeanParliamentary Elections in June were a particular focus. We published a blog1 illustratingthe steps taken by Snap to protect European democracy and the integrity of electoralprocesses, as well as the lessons learnt from the exercise. We have taken ourexperiences with these elections, as well as the Commission’s Guidelines into accountwhen assessing risks to democracy. For more information, see Section 4.3.1 Democracyand Electoral Processes in this Report. ● Minors - We have also seen intense discussion on the role of online platforms in the livesof young people. Over the last year, the world has witnessed CEOs of major technologycompanies in front of the US senate, the publication of draft codes of practice on 1 https://newsroom.snap.com/snap-eu-election. 12 CONFIDENTIAL protecting minors under the UK’s Online Safety Act and wide ranging debates in particularon screen time and age assurance. In the European Union, a new quantitative study fromthe Netherlands that compares online platforms on well-being, self-esteem and friendshipcloseness found that Snapchat was the only online platform studied that positivelyimpacts well-being and Snapchat also has a strong positive effect on friendships and nonet negative effect on self-esteem.2 We have been working proactively to support theCommission’s drive for an AADC code under the EU BIK+ strategy and the upcomingguidance for Article 28 of the DSA to provide a clear, harmonized direction to ensure ahigh level of privacy, safety and security across all online platforms. For more information,see Sections 4.2.6 (Children’s Rights), 4.4.3 (Negative effects on Children) and 5.8(Protection of Minors). Over the next year, we look forward to continuing our constructive dialogue with the Commission,our Digital Service Coordinator and other stakeholders to further the objectives of the DSA, aswell as encouraging continued close collaboration with those cover seeing the General DataProtection Regulation, Audio Visual Media Services Directive and the Digital Markets Act toensure a holistic approach to privacy, safety and security across these digital platform laws. 2 Social Media Use Leads to Negative Mental Health Outcomes for Most Adolescents, Amber van der Wal,Ine Beyens, Loes H. C. Janssen, and Patti M. Valkenburg, 2024, url (preprint) 13 CONFIDENTIAL What is New? Snap is required to complete a report every year setting out the results of its risk assessment anddetails of its mitigations pursuant to Article 42(4). Previous reports Our first report was completed in August 2023 (the “2023 Report”). This was sent to theCommission without undue delay as required by Article 42(4). Snap’s reports are published on itsEuropean Union Transparency page3 within 3 months of Snap having received its final auditorsreport (as this includes an audit of Snap’s compliance with its risk and mitigation assessment andreporting obligations) pursuant to Article 42(4) i.e. approximately 1 year and 3 months after eachreport is completed. What is new in this Report? Section 1 - Introduction ● Snapchat 101 \- In all material respects, Snapchat’s in-scope services remain the same since our 2023 Report. Our data shows that the vast majority of our users are stillprimarily using the messaging aspects of our platform, and we continue to believe this isan important lens through which to view Snapchat. We have not deployed functionalitiesthat were likely to have a critical impact on the risks identified pursuant to Article 34 ofthe DSA. ● Snapchat Community \- We continue to observe positive growth in our user base globally, and in the European Union we grew to 92.4 million average monthly active recipients ofour Snapchat app (as at 1 August 2024). Our community demographics have not seen anysignificant changes since our 2023 Report. Section 2 - DSA Risk Assessment Scope ● Scope Assessment \- Since the 2023 Report: (i) our Snapchat designation has not changed; (ii) the Commission has not issued any new guidance relating to scope and (ii)the functionality of Snapchat has not significantly changed. We have therefore confirmedthat Snap still considers the Spotlight, Discover, Public Profiles, Snap Map, Lenses, andAdvertising services of Snapchat to fall within the scope of our risk assessment andmitigation obligations in Articles 34 and 35. We have confirmed that Snap continues to 3 https://values.snap.com/privacy/transparency/european-union 14 CONFIDENTIAL consider My AI and other similar generative AI tools made available by Snap to be out ofscope of Snapchat’s designation except for one advertising case identified below. ● In our short descriptions of the in-scope aspects of Spotlight, Discover, Public Profiles, Snap Map, Lenses, and Advertising services of Snapchat, we have noted in particular that: ○ We have removed reference to Spotlight Challenges in Spotlight as this featurehas now been deprecated. ○ We have acknowledged that content created by generative AI tools (whether onthird party platforms or Snapchat) could be disseminated via Spotlight andDiscover and noted that Snap has taken this into account in this Report. ○ We have no significant updates, Snapchat+ features, or generative AI featuresrelevant to Public Profiles. ○ We have flagged the launch of simplified location sharing and two new Snapchat+features on the Snap Map, but confirmed that we have assessed these to be outof scope of Snapchat’s designation. ○ We have noted that certain Lenses are only available to Snapchat+ subscribersand flagged the launch of GenAI Suite in Lens Studio (but confirmed that we haveassessed the latter to be out of scope of Snapchat’s designation). ○ We have flagged that our ad creation tools incorporate common, minor tools togenerate assets such as text translations and background images. We haveassessed that these are likely to be considered in scope. Section 3 - DSA Risk Assessment Methodology ● DSA Risk Assessment Methodology \- We have confirmed that there is no change to our risk assessment methodology since our 2023 Report. We have added an additionalparagraph on how we deal with improvements identified in our risk assessment result andmitigation reports, and confirmed that we collate and retain supporting documentation fora minimum of 3 years. Section 4 - DSA Risk Assessment Results ● We have updated the likelihood, severity, overall potential risk prioritization assessments,mitigation assessments and confirmed that there have been no changes to theconclusions we reached in our 2023 Report that we have reasonable, proportionate andeffective mitigation measures for each systemic risk identified in Article 34 of the DSA: ○ Category 1 \- Dissemination of content that is illegal or violates our terms and conditions i.e. dissemination of CSAM, Hate Speech, Sale of prohibited Goods,Terrorist content, IP Infringement, Adult Sexual Content, Harassment and Bullying, 15 CONFIDENTIAL Self-Harm, Violent and Dangerous, Harmful False Information, Fraud and Spam,and Other illegal activities ○ Category 2 \- Negative Effects on Fundamental EU Rights i.e. Human Dignity, Freedom of Expression and Information, Private Life, Data Protection,Non-discrimination and Freedom of Religion, Children’s Rights, ConsumerProtection and Property. ○ Category 3 - Negative effect on democratic and electoral processes, civic discourse and public security ○ Category 4 - Negative Effects on Public Health, Gender-based Violence, Protection of Children and Physical and Mental well-being. ● We have seen a substantial reduction in prevalence rates across all of the illegal andother violating content categories that we monitor (see our update on prevalence testingbelow). As a result, we have been able to lower the relative likelihood of two of our riskcategories: ○ Our Adult Sexual Content risk has been lowered from highest relative likelihood to our medium relative likelihood category. ○ Our Fraud and Spam risk has been lowered from medium relative likelihood to our lowest relative likelihood category. We had identified both categories as a focus for our ongoing monitoring andmanagement of risk in the conclusion of our 2023 Report. We have noted that we willcontinue to monitor both categories, but are very pleased with the progress. ● The prevalence of violating content relating to regulated goods risk has also fallen fromour medium relative likelihood category to our lowest relative likelihood category, tomatch the prevalence of ‘illegal goods and activities’ (which was already in our lowest relative likelihood category). This reinforces our assessment that our Sale of prohibited Goods risk continues to fall within our lowest relative likelihood category. ● In our 2023 Report, we also identified Harassment \& bullying as focus for ongoing monitoring as it had recently seen a slight rise in prevalence and reporting. Since then, wehave observed a significant reduction in prevalence of this content on the inscopeservices of Snapchat and confirm that reporting and enforcement rates relating to theinscope services on Snapchat are very low. We have therefore noted that we will continueto monitor this category, but maintain our conclusion that we have reasonable,proportionate and effective measures. ● We have made significant revisions to our assessment relating to the risk of negative effects on democratic and electoral processes to reflect our recent positive experiences in the run up to and during the European elections and to address the Commission’sGuideline for providers of VLOPs and VLOSEs on the mitigation of systemic risks for 16 CONFIDENTIAL electoral processes.4 We continue to conclude that we have reasonable, proportionateand effective mitigation measures. ● We have also updated our assessment relating to the risk of negative consequences on physical and mental wellbeing, in particular to take account of a new quantitative study from the Netherlands (currently in pre-print) that compares online platforms on well-being,self-esteem and friendship closeness.5 This found that Snapchat was the only onlineplatform studied that positively impacts well-being and Snapchat also has a strongpositive effect on friendships and no net negative effect on self-esteem. ● General - We have also noted in the specific mitigation summary for each risk thefollowing general mitigation updates: ○ Regarding the measure that we have put in place with regards to the Protection ofMinors, we have updated each summary to confirm the launch of our new parentssite which provides additional guidance for parents and carers on risks andsupport.6 ○ Regarding the measures that we have put in place with regards to ContentAuthenticy, we have updated each summary to confirm that Snap has taken stepsto mitigate the risk that (i) its generative AI tools are used for illegal or otherwiseviolating content and (ii) illegal or otherwise violating content created usinggenerative AI tools on any online platform are disseminated on Snapchat’sinscope services. Section 5 - Specific Mitigations ● Snapchat Design and Function \- We have not noted any significant changes to Snapchat’s Design and Function mitigations. We have provided additional screenshots relating to ourlaunch of simplified location sharing on Snapchat, although this is out of scope of thisReport. We have added additional information on the process and governance around thedesign of Snapchat’s online interface. ● Terms \- We have significantly expanded this section to provide additional information regarding the oversight and administration, content, accessibility and readability of ourterms mitigations. 6 https://parents.snapchat.com. 5 Social Media Use Leads to Negative Mental Health Outcomes for Most Adolescents, Amber van der Wal,Ine Beyens, Loes H. C. Janssen, and Patti M. Valkenburg, 2024, url (preprint) 4 Guidelines for providers of VLOPs and VLOSEs on the mitigation of systemic risks for electoral processes,April 2024, url. 17 CONFIDENTIAL ● Transparency - We have not noted any significant changes to our transparency mitigations. We have provided some updates regarding updates to our Privacy Policy,additional safety resources added to our Safety Center, the launch of our dedicatedmicrosite for parents, educators and other caregivers. ● Moderation and Enforcement \- We have not noted any significant changes to our moderation and enforcement mitigations. We have made some presentationaladjustments to ensure the information provided is located in the right section, as well asupdating our statistics, providing more detail on oversight and administration, qualityassurance, notification of criminal offenses and complaint handling. ● Algorithmic Systems and Targeted Advertising \- We have not noted any significant changes to our algorithmic system and targeted advertising mitigations. We haveprovided some additional information regarding oversight and administration, modeldevelopment and deployment, change management and monitoring and qualityassurance. ● Protection of Minors \- We have provided updates on the mitigation measures we put in place to protect minors, including clarifications regarding our use of inferred agetechniques, readability of our terms and conditions, administration and oversight,overview as to how we apply the age appropriate design code, confirmation that we havedisabled targeted advertising to minors, updates to the Family Center and otherclarifications. ● Content Authenticity \- We have made significant updates to this section to highlight how we have accounted for advancements in generative AI technologies in our riskassessments and our approach and measures in place to mitigate these risks both inrespect to the use of generative AI in content creation (which we continue to consider tobe out of scope of Snap’s designation (save for certain commonplace ad creation tools)and are providing for context) and in dissemination to the public on Snapchat’s inscopeservices. ● Trusted flaggers \- We have provided additional information regarding Snap’s trusted flagger program, as updating our statistics and trend analysis. ● Dispute Settlement Bodies - We have provided minor updates regarding the establishment of out-of-court settlement bodies and our approach, including noting thatthe Commission has not yet published a list of certified bodies. ● Codes and Crisis Protocols \- We have provided updates regarding Snap’s engagement with the European Commission and other stakeholders on updates to the EU Hate 18 CONFIDENTIAL Speech Code, and providing a recent case study where we have exercised our contentcrisis management protocol. Section 6 - Ongoing Risk Detection \& Management ● Platform Principles Framework \- There have been no significant updates to our platform principles framework. ● DSA Compliance Team and Cross-Functional Working Groups \- We have provided some additional information regarding roles and responsibilities, our compliance officers,independent compliance function, DSA management body, DSA Cross-FunctionalGovernance Team, Points of Contact, Legal Representative and Supervisory Fee. ● Privacy and Safety by Design \- We have updated this section in particular to explain how we conduct our critical impact checks, as well as confirm that since our 2023 Report, wehave not identified any deployed functionalities that were likely to have a critical impacton our assessment of risks and mitigations pursuant to Articles 34 and 35 of the DSA. ● Prevalence Testing \- We are extremely pleased with the progress we have observed from our prevalence testing over the last year, which demonstrates that the effectiveness ofour proactive detection mechanisms, agent training and other content moderation andenforcement efforts has continued to increase significantly since our 2023 Report. Inparticular: ○ We have observed a significant decrease in our overall ‘Policy ViolatingPrevalence’ (PVP) rate. ○ All of the most prevalent violating content categories have significantly reducedPVP rates. ○ There are now no illegal or other violating content categories in our low likelihoodcategory (which is a PVP of 0.5% and above). ○ Child Sexual Exploitation and Abuse Imagery (CSEAI) is no longer categorizedamong the most significant sources of harm ● External Request Monitoring and Review \- We have confirmed that we continue to produce transparency reports and monitor advertising review rejections, advertisingreporting and enforcements, ‘privacy, data protection and DSA’ requests and generalcommunity support requests, and use this data to support the conclusions reached in thisReport. Since the DSA came into force, we have also been monitoring DSA queries raisedvia our dedicated email address and community support page. 19 CONFIDENTIAL ● Digital Well-Being Index (DWBI) Initiative \- We have updated this section to report on the results from the Year Three industry wide DWBI research study. ● Snap Advisory Groups \- We have provided information on the progress of our work with the Snap Safety Advisory Board and the establishment of a new Snap Council for DigitalWell-Being. ● Audit \- We have noted the completion of our external DSA audit of Snap’s compliance with its obligations under Chapter 3 of the Digital Services Act for the audit periodbetween August 25th 2023 and June 30th 2024 pursuant to Article 37, and that we willshare the audit report in due course pursuant to pursuant Article 42.4(c) and (d). Annex - Explainer Series ● Annex - Explainer Series \- There have been a number of minor updates to our Explainer Series and references to the new versions are included. Conclusion ● Conclusion - We note that we have carried out a risk assessment of Snapchat’s in-scope services and continued to confirm that we have in place reasonable,proportionate and effective mitigation measures, tailored to the specific systemic risksidentified. We have also reflected on the process with the areas for improvementidentified in our 2023 Report, as well as identifying new areas for improvement over thecoming year. 1\. Introduction At Snap, our mission is to contribute to human progress by empowering people to expressthemselves, live in the moment, learn about the world, and have fun together. Even as Snap grows and faces new opportunities and challenges, we remain grounded inkindness. Our engineers, designers, product managers, and other team members build ourproducts and services to serve people. The well-being of the community informs our decisionmaking, which in turn creates more value for our business over the long term.7 1.1 Snapchat 1.01 Snapchat is a communications app designed for people ages 13 and up, who primarily use it totalk with their close friends, similar to the ways they interact in real life. It’s similar to how older 7 See our Citizen Snap Report for more details. 20 CONFIDENTIAL generations use text messaging or their phone to stay in touch with friends and family. Since our2023 Report, our data shows that the vast majority of our users are still primarily using themessaging aspects of our platform. While the products detailed in this report and within scope ofthe DSA primarily revolve around our public content surfaces, our core use is a messaging app,which sets us apart from many other VLOPs, and we believe is an important lens through whichto view Snap and our platform. We purposely designed Snapchat differently from traditional social media. It doesn’t open to apublic news feed powered by an algorithm with likes and comments. Instead, Snapchat opens toa camera and has five tabs: Camera, Chat, Map, Stories, and Spotlight.8 Since our 2023 Report, we have not deployed functionalities that were likely to have a criticalimpact on the risks identified pursuant to Article 34 of the DSA. There are a number ofsignificant product changes that we are considering for the future, including in particular: ● We are considering simplifying the application from 5 to 3 tabs. The simplified versionwould still open to the Camera, but would move most friend related content left of theCamera to one tab (Chats, Snaps, Stories, and access point for the Map), and more publiccontent right of the Camera in a unified content feed. We are running tests this year toassess how it performs. This is primarily a cosmetic change, and it should not impact themitigations we have in place across our application, including content moderation. Wehave not assessed it to have a critical impact on the risks we have identified in this Report. ● We are also considering a Public Profiles experience that is suitable for 16-17 year olds,which we may start to roll out in select countries later this year. We have provisionallyassessed that this change would likely have a critical impact on the risks we haveidentified in this Report. This experience will contain additional mitigations, on top of theexisting Public Profiles mitigations. A roll out in the EU would be dependent on us firstfinalizing the risk assessment we have been carrying out pursuant to our DSA and otherlegal obligations. As these product changes develop further, we would be happy to brief the Commission andprovide a more detailed preview. The next section provides a reminder of our platform architecture. 8 View our Snapchat 101 video for more details. 21 CONFIDENTIAL Camera Snapchat opens into a camera, making it an easy and visual way for people to share what’s ontheir mind with the people that matter most to them. Snapchatters can Snap a quick video orphoto with our augmented reality Lenses to put fun and educational layers on the world, and getcreative by overlaying text, stickers, and more. Chat To the left of the Camera is Chat, where Snapchatters can talk with their friends and family usingtext and pictures. Chats will show when both friends are there at the same time. They’ll alsoindicate when a friend has opened and viewed a Snap. Snaps and Chats delete-by-default to mirror real life conversations, where what one says or doesisn't recorded forever and shared with a bunch of strangers. This helps people feel morecomfortable expressing themselves, the same way they would if they were just hanging out withfriends in person. While Chats and Snaps delete by default, Snapchatters do have the option tosave Chats – simply by tapping on the ones they want to save. In Chat, you can also make voice and video calls and join group conversations and chat with MyAI, our chatbot powered by OpenAI’s ChatGPT technology. Map Swipe to the left of Chat for the Map. Our Map is an interactive way for Snapchatters to sharetheir favorite spots, discover new places, and see what their friends are up to – but only if theychoose to share their location with their friends. Profile 22 CONFIDENTIAL My Profile features a user’s Snapchat info, like their Bitmoji (which is an avatar representation of the user), location on the Map, friend info, and more. My Profile is also where Snapchatters canmanage their friendships, and report, block, or remove a friend. Public Profile Public Profiles enable Snapchatters to be discovered in the app. If Snapchatters want a PublicProfile, they will need to create one first. Once they have created a Public Profile, they canshowcase their favorite public Snaps and share Lenses and other information. Discover Swipe to the right of the Camera for Stories. Snapchatters can add Snaps to their Stories to sharemore of their day with friends and family, and scroll down to discover new Stories and contentabout the world — produced by trusted media publishers and popular creators. Spotlight Right next to Stories is our entertainment platform, Spotlight. This is where Snapchatters cansubmit and watch short, fun, and creative videos for our community. In Section 2 of this report we provide more details on the products and services that are in scopeof the DSA Risk Assessment. 1.2 Snapchat Community We reach over 850 million9 monthly active users around the world, and see a path for Snapchatto reach over 1 billion people in the next 2-3 years at our current growth rate. Additionally, wehave over 432 million10 daily active users globally. We provide information on the average monthly active recipients of our Snapchat app, across the EU and per EU Member State, in our European Union transparency page on our website. Our European Snapchatter community consists of a diverse range of ages and genders. WhileSnapchat does have a young demographic, by far the largest age category is 18-24, the secondlargest age category is 25-34, 35+ makes up the third place, and 13-17 is the smallest agecategory. 10 Snap Inc. public data Q2 2024, see https://investor.snap.com. 9 Snap Inc. public data Q2 2024, see https://investor.snap.com. 23 CONFIDENTIAL In terms of gender, our analysis indicates that our community is fairly balanced but with a slightlyhigher percentage of the community identifying as female. This has remained consistent sinceour 2023 Report. A more detailed analysis of gender shows a slightly higher percentage of our 18-24 age groupidentifying as male, with a higher percentage of our 35+ age group identifying as female. 24 CONFIDENTIAL 2\. DSA Risk Assessment Scope 2.1 Approach Articles 34 and 35 apply to Very Large Online Platforms designated by the EuropeanCommission. Snapchat was designated as a Very Large Online Platform by the Commission on25 April 2023 because the Average Monthly Active Recipients of Snapchat exceeds 45 million. The Commission Decision to designate Snapchat as a Very Large Online Platform states that itonly applies to services provided as part of Snapchat that meet the definition of online platformlaid down in Article 3, point (i), of Regulation (EU) 2022/2065. The designation does not apply toservices that are provided together with Snapchat, such as a private messaging service, and that,based on their technical functionalities, do not in themselves meet the definition of onlineplatform laid down in Article 3, point (i), of Regulation (EU) 2022/2065. Article 3.(i) of the DSA defines ‘online platform’ as: “a hosting service that, at the request of a recipient of the service, stores and disseminatesinformation to the public, unless that activity is a minor and purely ancillary feature of anotherservice or a minor functionality of the principal service and, for objective and technical reasons,cannot be used without that other service, and the integration of the feature or functionality intothe other service is not a means to circumvent the applicability of this Regulation”. Recital 14 explains that: “The concept of ‘dissemination to the public’, as used in this Regulation, should entail the makingavailable of information to a potentially unlimited number of persons, meaning making theinformation easily accessible to recipients of the service in general without further action by therecipient of the service providing the information being required, irrespective of whether thosepersons actually access the information in question. Accordingly, where access to information requires registration or admittance to a group ofrecipients of the service, that information should be considered to be disseminated to the publiconly where recipients of the service seeking to access the information are automaticallyregistered or admitted without a human decision or selection of whom to grant access.Interpersonal communication services, as defined in Directive (EU) 2018/1972 of the EuropeanParliament and of the Council,11 such as emails or private messaging services, fall outside thescope of the definition of online platforms as they are used for interpersonal communicationbetween a finite number of persons determined by the sender of the communication. 11 url. 25 CONFIDENTIAL However, the obligations set out in this Regulation for providers of online platforms may apply toservices that allow the making available of information to a potentially unlimited number ofrecipients, not determined by the sender of the communication, such as through public groups oropen channels. Information should be considered disseminated to the public within the meaningof this Regulation only where that dissemination occurs upon the direct request by the recipientof the service that provided the information.” 2.1.1 Scope in our 2023 Report Taking account of the above DSA definition and guidance, and that fact that Snapchatters areautomatically registered without a human decision or selection of whom to grant access, for our2023 Report, Snap considered the Spotlight, Discover, Lenses, Public Profiles, Snap Map, andAdvertising services of Snapchat to fall within the scope of risk assessment and mitigationobligations in Articles 34 and 35. These services entailed making information published byrecipients of those services easily accessible to other recipients of Snapchat in general withoutfurther action by the recipients publishing the information in question. 2.1.2 Scope for this Report For this Report, (i) the DSA definitions and guidance remain the same and (ii) the functionality ofSnapchat has not significantly changed and Snapchatters are still automatically registeredwithout a human decision or selection of whom to grant access. Therefore, Snap still considersthe Spotlight, Discover, Lenses, Public Profiles, Snap Map, and Advertising services ofSnapchat to fall within the scope of risk assessment and mitigation obligations in Articles 34and 35. These services still entail making information published by recipients of those serviceseasily accessible to other recipients of Snapchat in general without further action by therecipients publishing the information in question. When we refer to “Snapchat” or “Snapchat’s in-scope services” in this Report, therefore, we arereferring to those six services in Snapchat unless the context is clear that it is referring toSnapchat as a whole. The six in scope services of Snapchat are described in more detail in the following sections. 2.2 Spotlight What is Spotlight? Spotlight is the Snapchat community’s destination for entertaining short-form video content.Launched in November 2020, Spotlight provides users a simple way to view short-form videoscreated and submitted by the Snapchat community via a personalized feed. All users can postvideos to Spotlight either via the Snapchat app or on the website, and videos on Spotlight arepublic and visible to users on the Snapchat app, on the web, and a link to the Spotlight video canbe shared to other platforms. Users can also add Comments to Spotlight videos, which go 26 CONFIDENTIAL through moderation before being shown to the creator to either accept or reject, orauto-approve. If accepted or auto-approved, the Comment is publicly visible on the Spotlightvideo. Spotlight Comments may be deleted or reported, and viewers can also indicate fondnessby clicking on a heart icon. In addition to compliance with Spotlight Terms, users must also comply with the Community Guidelines and the Spotlight Guidelines. Certain higher profile Snapchatters have the opportunity to receive revenue from their content if they, and their Spotlight Snaps, meet certain eligibility criteria during the Eligibility Period. How does Spotlight work? Spotlight provides a content experience that is intended to entertain and delight users in thesame app they use to communicate with their friends and family. It offers creators at all stages oftheir career a variety of opportunities and tools to help them grow their audiences, buildsustainable businesses and make content creation a full-time career. Spotlight is an easy entrypoint to start your creator journey and is a source of relevant cultural trends and credible partnerto the industry (media, music, sports, fashion, etc.) that offers meaningful reach, relevance andrevenue. The content shown in Spotlight is personalized to provide the user with a more relevant experience. Spotlight’s ranking algorithm is described here. Users may opt out of personalization as described here. Spotlight content is moderated using a combination of auto-moderation and human moderation, and all Spotlight content is human moderated before being widelydistributed. Spotlight also uses various engagement and metadata to determine eligibility toreceive revenue from their content. Snapchat+ Features relevant to Spotlight None. Gen AI Features being used by Spotlight None. Snap recognises that content created by generative AI tools (whether on third partyplatforms or Snapchat) could be disseminated via Spotlight. Snap has taken this into account inthis Report, including explaining how this activity impacts our risk assessment and what measureswe have taken to address risks relating to the dissemination of content created by generative AI. 2.3 Discover What is Discover? Discover is part of the 4th tab in the Snapchat app, below your friends’ Stories. Contrary to our2023 Report, Snap ultimately decided not to rename this product from Discover to For You.Hence, any reference to For You that remains in this Report or in supporting documentationshould be interpreted as a reference to Discover. 27 CONFIDENTIAL Discover is dedicated to Creator Stories, which includes Media Partner content, and some usergenerated content (“UGC”) created from Snaps by popular users (“Creator Content”). The UGCthat appears on Discover includes the Public Stories from Snap Stars and other users who meet afollower count threshold. The videos in Discover are accessible to all users including thosebetween 13-17 years old. How does Discover work? Discover displays personalized content to users. Discover achieves this using its ranking algorithm, which is described here. The intended purpose of this processing is to personalize Discover and make it easy for users to discover new content that is relevant to their interests. Theintended effect/impact on users is that they enjoy what they are watching and remain engaged users of Snapchat. Users may opt out of personalization as described here. Discover also generates information about how Snapchatters interact with the content inDiscover. It achieves this by generating ‘event’ metadata each time a user does somethingnoteworthy, like viewing or skipping a video. The intended purpose of this processing is to selectcontent the user is likely to be interested in, in order to further personalize content on Discoverand elsewhere in Snapchat (such as other content areas like Spotlight and also Advertising - therevenue from which is used to pay for Snapchat). The intended effect/impact on users is that theyenjoy their experience and remain engaged users of Snapchat. Snapchat+ Features relevant to Discover None. Gen AI Features being used by Discover None. Snap recognises that content created by generative AI tools (whether on third partyplatforms or Snapchat) could be disseminated via Discover. Snap has taken this into account inthis Report, including explaining how this activity impacts our risk assessment and what measureswe have taken to address risks relating to the dissemination of content created by generative AI. 2.4 Public Profiles What are Public Profiles? Public Profiles enable Snapchatters to be discovered and followed in the app and showcase theirfavorite public Snaps, Lenses and other information. Snapchatters (including businesses) cancreate and access Public Profiles and grow an audience with their public identity. Public Profilesenables Snapchatters to showcase Stories, Spotlights and Lenses. For more information, see here. How do Public Profiles work? 28 CONFIDENTIAL Currently, Snapchatter accounts aged 18 and over can opt into having a Public Profile ifSnapchatters want to share a bit more about themselves with a wider audience (beyond theirimmediate friends). Creating a public profile is straightforward. An eligible Snapchatter is required to: (i) tap theirBitmoji or Story icon at the top to go to My Profile; (ii) Scroll down to the 'Public Profile' sectionand Tap 'Create Public Profile' and (iii) and then follow the simple instructions to create theirPublic Profile. With a public profile, a Snappchatter can: ● Add a Photo, Bio, Description, Location, Stories, Spotlights, and Lenses to your PublicProfile ● Be Followed by other Snapchatters ● Show their Follower Count ● View Public Story, Lens, and Audience Insights ● Add Snaps to their Public Story Snapchatters with a public profile that are particularly active can have their accounts upgraded toa Creator Account. These have advanced features that are designed to enable professionalCreators to connect and grow with their audience. Creator Accounts are eligible to have theircontent shown in the For You section of Snapchat. Snapchat+ Features relevant to Public Profiles None. Gen AI Features being used by Public Profiles None. 2.5 Snap Map What is Snap Map? Snap Map is designed to open up a world of possibilities for our community, enabling friends toexperience something new in the world every day. Through an interactive map interface, SnapMap shows users what’s happening nearby and around the world, anchored by the context offriends’ Bitmojis. It’s a personal map that starts with the user at the center and reflects the people,places, and activities they care about, and helps users meet up with friends, express themselves,find things to do, and explore places elsewhere. The Snap Map was developed with the privacyand safety of our community of Snapchatters in mind. How does Snap Map work? Snapchatters can share their Snaps to the Map by selecting “Snap Map” on the “Send To” page. Ifthe Snapchatter has a Public Profile or is sharing their My Story with everyone, they may also 29 CONFIDENTIAL have their Snap shared on Snap Map when it's tagged to a place or venue. Snapchatters can alsochoose to share their location on the Map with friends while the Snapchat app is actively beingused, or share their live location with them even when the app is backgrounded. Since our 2023 Report, we have simplified how location sharing works, which is highlighted in our blog post here. As location sharing is with Friends, rather than the public, and out of scope of Snapchat’sdesignation, we have not detailed this change in the Report. Nevertheless, we have outlinedsafeguards we have implemented for location sharing in the Mitigations section below (Sections5.1 and 5.3 specifically). Snap Map features five types of user-generated content that can be served: 1. Map Stories include thumbnails on the map that highlight interesting events and popularplaces on the Map. 2. Place Stories appear on Place profiles. They contain Public Stories snaps explicitlytagged with the place, using either venue filters or place stickers. 30 CONFIDENTIAL 3. City Stories appear in the header of the Map and display the best snaps in that localityfrom the last 7 days. They can appear for cities and neighborhoods. 31 CONFIDENTIAL 4. The Heat Map is used to visualize the volume and recency of content that’s submitted toPublic Stories. Content up to seven days old can be accessed by the heatmap. Heat spotsrepresent areas where there is recent, high volume content. 5. Friend Stories tagged with Places presents a view of snaps that have been tagged withPlaces by a user’s friend, along with the Bitmoji of the friend, that would appear with thePlace on the basemap. This helps to personalize places on the user’s Map, highlightingthe places friends have recently visited. Snap Map submissions may be stored for a while and may be visible on Snapchat for longperiods of time as explained to Snapchatters in privacy notices. Snapchatters can remove a Snapthey submitted to Snap Map or place-tagged in Spotlight at any time via their profile. 32 CONFIDENTIAL Snapchat+ Features relevant to Snap Map Customizations - Snap Map offers a couple of customization features for Snapchat+ usersincluding the ability to customize your home on the Map and adding your pet on the Map (eithervia a preselected group of pets or by creating a generative AI version of your pet). Footsteps - Snap Map also lets SC+ users see how much of the world they’ve explored. Thiscontent is only made available to the user, and we have assessed it to be out of scope ofSnapchat’s designation. Gen AI Features being used by Snap Map 33 CONFIDENTIAL Gen AI Pet - Snap Map offers a Gen AI pet feature as described above. Note however that whilethis feature is integrated into the Snap Map, the content is only available to Friends with whomthe user chooses to share their location. It does not disseminate content to the broader public,and we have assessed it to be out of scope of Snapchat’s designation. Nevertheless we haveoutlined our approach to both Gen AI creation and dissemination in the Mitigations section (inparticular Section 5.9 (Content Authenticity)). 2.6 Lenses What are Lenses? Snapchat Lenses are augmented reality (“AR”) experiences designed to transform the way users look and the world around them. Snapchatters frequently use Lenses for entertainment purposes,for example by creating Snaps with added 3D effects, digital objects, characters, andtransformations to their image and voice. For example, Lenses can be used to add a layer ofmake-up to the user’s face, to distort the user’s face, to add a different background or certain elements to the surroundings. The most popular Lenses at the moment can be found here. Snapchatters can interact with Lenses in the Carousel, via Search, and via Lens Explorer. In addition, we offer advertisers the possibility of creating Sponsored fes. How do Lenses work? Lenses (in popular language often dubbed as ‘filters’) are created by a relatively limited numberof community developers, and Snap’s internal Lens Team. The transformational effects of Lensesare often accomplished through object detection, which is an algorithm designed to help acomputer generally understand what objects are in an image. For example, it lets us know that anose is a nose or an eye is an eye. There are numerous AR development tools Snap has madepublicly available through Lens Studio, Snaps’ Lens development platform and there are alsointernal tools that only the Lens Team can use to develop Lenses. Snap’s AR development toolsare reviewed by privacy engineering and legal before being used in Lenses. Some examples ofAR development tools are object detection, text to speech, location landmarks and ML modelsand algorithms to support AR effects like tools for depth and context understanding, all designedto help a computer generally understand what objects are in an image. We provide provided further information about how Lenses works in product specific supportpages: ● How to use lenses ● Create Your Own Filters \& Lenses • Snapchat Snapchatters can create or develop Lenses in the desktop application ‘Lens Studio’. There is aPublic version and an internal Snap version of Lens Studio. Lens developers may publish Lensesthrough ‘My Lenses’, a web based portal. Lenses built by Snap’s Lens Team are organic Lenses. Snapchat+ Features relevant to Lenses 34 CONFIDENTIAL Certain Lenses are only available to Snapchat+ subscribers. Gen AI Features being used by Lenses Lens Studio features a GenAI Suite which lets developers take advantage of our generative AItechnology to create assets (such as text, effects and backgrounds) for Lenses. Note howeverthat Lens studio is only a creation tool. It does not disseminate content to the broader public, andwe have assessed it to be out of scope of Snapchat’s designation. Nevertheless we haveoutlined our approach to both Gen AI creation and dissemination in the Mitigations section (inparticular Section 5.9 (Content Authenticity)). 2.7 Advertising What is Snap’s Advertising product? Snap relies on online advertising to support its business. Snap Advertising is a digital ad productcreated for advertisers who would like to easily create and manage ads that target relevantaudiences on Snapchat. We process user information about Snapchatters to serve them with adswithin Snapchat that we think they might be interested in. An overview of Snap’s ads services can be found here and here. Some of Snap’s advertising tools allow advertisers to provide Snap with data about their customers to improve theiradvertising campaigns. These tools are explained here: ● Custom List Audiences ● Snap Pixel ● Conversion API ● Advanced and Estimated Conversion In addition, we offer advertisers the possibility of creating Sponsored Lenses. How does Advertising Work? Our ad ranking algorithm determines which ads are displayed to a Snapchatter who is in theselected audience for those ads. The ad ranking algorithm uses various signals, including prior adinteractions and social signals, to determine which ads that user is more likely to interact with andthen combines this with the results of advertiser ad action for that Snapchatter, to select an ad todisplay. Snap analyzes prior ad interactions to target advertisements. For example, we maydetermine that a user is likely to swipe up on certain types of ads or download certain types ofgames when they see an ad on Snapchat. We may then use this information to show that usersimilar ads. Snapchatter interactions with the ad (i.e. impression data) is then logged to (a) attributeimpressions to conversion events (such as a purchase on an advertiser website or download ofan advertiser app) to demonstrate the performance of the ad and (b) to further train the adranking algorithm. 35 CONFIDENTIAL Snapchat+ Features relevant to Advertising None Gen AI Features being used by Advertising Advertisers can take advantage of generative AI tools during ad creation (such as texttranslations and background image creation). These are in systems directly related to thedissemination of content (advertising) to the public on Snapchat (although they are common toolsand minor in nature). We have taken the use of generative AI tools into account in this Report (inparticular Section 4.3.1 (Democracy/Elections) and Section 5.9 (Content Authenticity)). 36 CONFIDENTIAL 3\. DSA Risk Assessment Methodology There have been no changes to our risk assessment methodology since our 2023 Report. Inorder to meet its obligations under Articles 34 and 35 of the DSA, Snap has applied a standardrisk methodology adapted from that commonly used to assess risks in other contexts, including the EU general risk assessment methodology for product safety12 and the ICO’s DPIA template. It is worth noting that, in the UK, Ofcom’s recent consultations on its illegal harms and minors codesof practice have proposed a similar methodology in its draft risk assessment guidancedocuments.13 The risk assessment methodology used by Snap has several steps: 3.1 Identification of Risks As a first step, Snap identified potential systemic risks for each of the four categories outlined inthe DSA: a. Category 1 (Article 34.1.(a) / DSA Recital 80): Dissemination of illegal or violating content,particularly rapidly and widely or as a result of intentional / automated manipulation,including: i. child sexual abuse material ii. illegal hate speech iii. criminal offenses and the conduct of illegal activities, such as the sale ofprohibited products or services, dangerous or counterfeit products, orillegally-traded animals b. Category 2 (Article 34.1.(b) / DSA Recital 81): Impact on fundamental EU rights, includingin particular rights for: i. Human dignity ii. Freedom of expression and of information, including media freedom and pluralism iii. Private life iv. Data protection v. Non-discrimination vi. Children vii. Consumer protection c. Category 3 (Article 34.1.(a) / DSA Recital 82): Negative effects on: i. Democratic and electoral processes ii. Civic discourse iii. Public security 13 Consultation on online harms, url and consultation on protecting children from online harms, url. 12 EU general risk assessment methodology (Action 5 of Multi-Annual Action Plan for the surveillance ofproducts in the EU (COM(2013)76)), url. 37 CONFIDENTIAL d. Category 4 (Article 34.1.(a) / DSA Recital 83): Negative effects, in particular from designand use/misuse such as a coordinated disinformation campaign, on: i. Public health ii. Gender-based violence iii. Children iv. Physical and mental well-being (including addictions) 3.2 Likelihood Analysis As a second step, Snap analyzed the extent to which the identified risk(s) are likely to occur onSnapchat. In practice the prevalence of almost all of Snapchat’s risks are considered to be verylow, in part because of robust mitigations and the inherent design of relevant Snapchatfunctionality, so Snap used a measure of relative likelihood between each risk on Snapchat sowe can continue to prioritize and improve (as explained in the following table). Note: this is notmeasuring likelihood relative to other platforms; it is measuring likelihood relative to risksassessed by Snap. With this in mind, Snap used three levels of relative likelihood: Relative likelihood of risk occurring on Snapchat Description Low likelihood This means this risk has the highest chance of occurring on Snapchat vs other risks. Where Prevalence Testing data is available, this risk has a percent of policy-violating prevalence (PVP) of 0.5% - 1,5%. Very low likelihood This means this risk has an average chance of occurring on Snapchat vs other risks. Where Prevalence Testing data is available, this risk has a percent of policy-violating prevalence (PVP) of between 0.05% and 0.49%. Extremely low likelihood This means this risk has the lowest chance of occurring on Snapchat vs other risks. Where Prevalence Testing data is available, this risk has a percent of policy-violating prevalence (PVP) of 0.049% or less. In order to assess likelihood, Snap uses a mix of internal information (such as Prevalence Testing data or illegal / violating content reporting data or complaint data, input from our safety advisoryboard and Snap commissioned research) and external information (such as external research,news reports and government and NGO guidance). Where internal information is required, thiswas obtained from the relevant teams responsible for maintaining that information. 38 CONFIDENTIAL 3.3 Severity Analysis As a third step, Snap analyzed the severity of the identified risk(s) by considering evidence of thepotential harm they have caused individuals or society in general. In practice the severity of allthe identified risks could cause at least significant harm (which is why they have beenidentified). So we used a measure of relative severity between each risk so it can continue toprioritize and improve. With this in mind, Snap used three levels of severity: Harm classification industry wide Description Severe harm industry wide This means this risk has the highest severity vs other risks. We consider severe harm to include both (1) harms that risk significant damage to the physical or emotional well-being of Snapchatters and society at large e.g. external parties influenced by (other people’s use of) Snapchat, and (2) the imminent, credible risk of severe harm, including threats to human life, safety, and well-being. Serious harm industry wide This risk has a medium level of severity vs other risks. We consider these risks not to be severe (as defined above) but still have the potential to cause serious harm. Significant harm industry wide This means this risk has the lowest severity vs other risks. While not the most severe or serious, these risks still have the potential to cause significant harm. The safety of Snapchatters is our top priority. We take behavior that threatens the safety of ourcommunity very seriously. We collaborate with experts, safety groups, and law enforcement onthese topics in order to better educate ourselves and our community, and to ensure we aresufficiently informed to analyze different levels of severity for each risk. 3.4 Overall Potential Risk Prioritization Assessment As a fourth step, Snap confirmed an overall potential risk prioritization for each identified risktaking account of the likelihood and severity analysis outlined above. This prioritization helps usto assess whether the mitigations we have put in place (as described in Snap's Mitigations) areproportionate, reasonable and effective as required by Article 35. As a guide we use thefollowing matrix that is commonly used in risk assessment methodologies to determine theoverall potential risk. However, this is only an approximation and we make a decision on theoverall potential risk, and therefore the prioritization, of a particular issue on a case by case basisdepending on the harm classification industry wide or the relative likelihood of risk occurring onSnapchat. As a result, there are instances where we deviated from the overall potential risk 39 CONFIDENTIAL prioritization matrix below (and we have explained each of these deviations in the relevantsub-sections of Section 4 - DSA Risk Assessment Results). Overall Potential Risk Prioritization Matrix Harm classificati on industry wide Severe harm industry wide Level 3 Level 1 Level 1 Serious harm industry wide Level 3 Level 2 Level 1 Significant harm industry wide Level 3 Level 3 Level 3 Extremely low Very low Low Relative likelihood of risk occurring on Snapchat 3.5 Snap's Mitigations As a fifth step, Snap considered the mitigation measures that it has taken to address each of therisks identified in the overall potential risk prioritization assessment. When considering thesemitigations, Snap has taken into account in particular the list of possible mitigations set out inArticle 35.1. For ease of reference, we have set out a table below that maps the Article 35.1 list ofmitigations to the corresponding section of this report where Snap has explained how it is usingthat mitigation measure on Snapchat. # DSA Mitigation Relevant Report Section a Adapting the design, features or functioning of their services, including their online interfaces. Snapchat Design and Function b Adapting their terms and conditions and their enforcement. Terms and Enforcement c Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Moderation d Testing and adapting their algorithmic systems, including their recommender systems. Algorithmic Systems 40 CONFIDENTIAL e Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Advertising Systems f Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Risk Detection and Management g Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Trusted Flaggers h Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Codes and Crisis Protocols i Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Transparency j Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate; Protection of Minors k Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Content Authenticity 3.6 Conclusions As a final step, Snap confirmed whether the mitigation measures it has taken were reasonable,proportionate and effective for each risk identified. To determine this, Snap considered if themitigations it has in place were effective to address the risk, given its overall potential risk prioritization category, by considering available evidence from its Prevalence Testing data or illegal / violating content reporting data or complaint data, input from our safety advisory board 41 CONFIDENTIAL and Snap commissioned research) and external information (such as external research, newsreports and government and NGO guidance). Where there was evidence that the existing measures risks may need some improvement toensure reasonable, proportionate and effective measures had been taken, Snap identified this inits conclusion and explained what steps it would be taking to achieve the improvement. 3.7 Supporting Documentation The data and documentation supporting the risk and mitigation assessment report werepreserved and will be retained for a minimum of 3 years. 42 CONFIDENTIAL 4\. DSA Risk Assessment Results In this Section of the Report, we explain the result of the risk assessment of Snapchat’s in-scopeservices that Snap has carried out pursuant to Article 34 of the DSA. This risk assessment wasconducted in accordance with the scope and methodology explained in Section 1 of this Report.One general point to note is that these risks impact a wide range of individuals, including ourSnapchatter community, victims of crime, the general public and the moderators that review thecontent on Snapchat. The results of this risk assessment apply to all such individuals, and whereappropriate we have noted impacts that extend beyond Snapchat (including the wellness of ourmoderators). It is Snap’s mission to reduce virtually all harmful content on our platform. To that end, we arecontinually improving our systems every single day, and are investing into (machine learning)technology, human moderation, and other measures to make our platform safer for our community. As described in the Ongoing Risk Management section below, Snap has reasonable, proportionate and effective measures to detect and manage risks on an ongoing basis. 4.1 Category 1 - Dissemination of content that is illegal orviolates our terms and conditions (Article 34.1.a / DSA Recital 80) In this first part we report on our assessment of the risk of illegal content or content that is incompatible with our Terms being disseminated on Snapchat as required by Article 34.1.a (“Category 1”), including in particular the illegal content identified in Recital 80. In our assessment,we have taken account of the extent to which these risks are influenced by intentionalmanipulation, including by inauthentic use or exploitation of the service, as well as the extent towhich Snapchat allows for amplification and potentially rapid and wide dissemination. The table below provides a summary of the results of our assessment of likelihood, severity andoverall potential risk prioritization, together with our conclusions given the mitigations that Snaphas put in place for each Category 1 risk. Category 1 - Dissemination of content that is illegal or violates our terms and conditions (including our Community Guidelines) Category Relative likelihood of risk occurring on Snapchat Harm classificatio n industry wide Risk Prioritization Conclusion 43 CONFIDENTIAL 4.1.1 Dissemination of child sexual abuse material Extremely low Likelihood Severe harm industry wide Level 1 Low Risk / Reasonable, proportionate and effective mitigations 4.1.2 Dissemination of illegal hate speech Extremely Low Likelihood Significant harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective mitigations 4.1.3 Dissemination of information related to the sale of prohibited products or services (such as dangerous products, counterfeit products or illegally-traded animals) Extremely Low Likelihood Severe harm industry wide (Drugs) Level 1 (Drugs) Low Risk / Reasonable, proportionate and effective mitigations Extremely Low Likelihood Serious harm industry wide (Weapons) Level 2 (Weapons) Low Risk / Reasonable, proportionate and effective mitigations Extremely Low Likelihood Significant harm industry wide (Other goods) Level 3 (Other goods) Low Risk / Reasonable, proportionate and effective mitigations 4.1.4 Dissemination of terrorist content Extremely Low Likelihood Serious harm industry wide Level 2 Low Risk / Reasonable, proportionate and effective mitigations with monitoring due to a very slight increase in prevalence to ensure this remains very low. 4.1.5 Dissemination of content that infringes on intellectual property rights Extremely Low Likelihood Significant harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective mitigations 4.1.6 Dissemination of adult sexual content Extremely Low Likelihood (Adult sexual crimes) Serious harm industry wide (Adult sexual crimes) Level 2 (Adult Sexual Crimes) Low Risk / Reasonable, proportionate and effective mitigations Very Low Likelihood (Other adult sexual content) Significant harm industry wide (Other adult sexual content) Level 3 (Other adult sexual content) Reasonable, proportionate and effective mitigations, which are being monitored to confirm prevalence continues to decline and further measures are not required. 4.1.7 Dissemination of content regarding harassment \& bullying Very Low Likelihood Serious harm industry wide Level 2 Low Risk / Reasonable, proportionate and effective mitigations, with ongoing monitoring. 44 CONFIDENTIAL 4.1.8 Dissemination of content that glorifies self-harm, including the promotion of self-injury, suicide or eating disorders Extremely Low Likelihood Serious harm industry wide Level 2 Low Risk / Reasonable, proportionate and effective mitigations 4.1.9 Dissemination of content encouraging or engaging in violent or dangerous behavior Extremely Low Likelihood Significant harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective mitigations 4.1.10 Dissemination of harmful false misinformation Extremely Low Likelihood Significant harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective mitigations 4.1.11 Dissemination of fraud and spam Low Likelihood (Advertisin g) Significant harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective mitigations, which are being monitored to confirm prevalence continues to decline and further measures are not required. Extremely Low Likelihood (Content) 4.1.12 Dissemination of information related to other illegal activities Extremely Low Likelihood Significant harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective mitigations, which we will keep under review. 4.1.1 Dissemination of child sexual abuse material Snap has recognised the risk of dissemination of child sexual abuse material (CSAM) on internetplatforms and services, including on Snapchat, for some time. Without mitigations, CSAM canconceivably appear in any of Snapchat’s in-scope services displaying user generated content,from videos featured on Spotlight / Discover to Lenses being used to add a Lens on top of CSAMcontent or upload an image containing CSAM (elements) in the Lens creation flow. Note,internally and externally, Snap uses the term Child Sexual Exploitation and Abuse Imagery(CSEAI) to refer to CSAM. Throughout this report, we will largely be using CSEAI to refer toCSAM. Likelihood , Snap measures Policy Violating Prevalence (PVP) via random sampling of Public Stories toestimate the percent of policy-violating views.14 All of the risks we track on Snapchat have a lowprevalence compared to the prevalence of these issues elsewhere online and offline. To aid ourprioritization, our methodology seeks to assess the relative likelihood between the risks we track 14 See Section 6.4 (Prevalence Testing). 45 CONFIDENTIAL (even though all are low in absolute terms). In our 2023 Report, we noted that CSEAI contentrepresented an extremely low percentage of total views of Snaps in Public Stories. We also notedthat in the second half of 2022, the proactive moderation detected and actioned 94% of the totalchild sexual exploitation and abuse violations reported in our Transparency Reports. In February 2024 by prominent Finnish NGO Protect Children published a study on the use oftech platforms by online child sexual abuse offenders, funded by the Tech Coalition and SafeOnline.15 Snapchat ranks last among the social media platforms used to search, view, and shareCSAM (10%), compared to Instagram (29%), Twitter/X (26%), Discord (23%)TikTok (21%), Facebook(20%), Youtube (18\&), Reddit (17%). As at 30 July 2024 and as a result of our continued focus, the percentage of CSEAI violatingviews has seen a further, substantial fall to an extremely low rate. The steps Snap has taken tomitigate this risk have diminished the likelihood that Snapchatters will encounter CSEAI onSnapchat’s in-scope services. Although Snap is aware of concerns regarding the dissemination ofCSEAI created using generative AI tools in the wider industry, Snap has not identified this asbeing a material issue on Snapchat’s in-scope services. As a result, Snap continues to place CSEAI into the Extremely low Likelihood. Severity For this Report, as in our 2023 Report, Snap still considers all CSEAI to have a risk of the ‘severeharm’ category To this end, Snap has assessed information published by governments and otherthird party sources. Overall potential risk prioritization Although the prevalence of CSEAI on Snapchat has continued to decline and is now at extremelylow levels (and is considered to be at the lowest level of all our risks), due to the potential for themost severe harms to be caused by CSEAI and the continued growth in the prevalence of thisissue online in general, Snap still considers CSEAI to be a Level 1 risk priority. There is nochange in this assessment from our 2023 Report. As described in our risk methodology section, we assess overall potential risk on a case by casebasis and Snap reserves the option to deviate from the overall potential prioritization risk matrixwe use as a guide. This is one of the cases where we have chosen to deviate. Snap's Mitigations Highlights 15 Tech Platforms Used by Online Child Sexual Abuse Offenders, February 2024, url. 46 CONFIDENTIAL Snap prohibits any activity that involves sexual exploitation or abuse of a minor, including sharingCSEAI, grooming, or sexual extortion (sextortion). By using Snapchat, users agree under our Terms not to post, save, send, forward, distribute, or ask for nude or sexually explicit content involving anyone under the age of 18 (this includes sending or saving such images ofthemselves). It is possible, despite Snap’s policies and enforcement efforts, that malicious actors will find waysto circumvent Snap’s enforcement mechanisms and practices in order to post CSEAI, which couldthen appear on Snap’s public surfaces. Preventing and addressing potential CSEAI is a top priority for Snap, and is considered a “severe harm” under Snap’s Community Guidelines, and we respond with swift and strict consequences against violators as explained in our Severe Harms explainer. As explained in the Moderation section (specifically, the section on CSEAI), we also proactively scan all Stories and Spotlight submissions using PhotoDNA and Google CSAI Match, and enforceagainst accounts found to be sending CSEAI. Snapchatters can also report CSEAI to us via in-appreporting options and anyone can submit a report through the Snapchat Support Site. When Snap becomes aware that CSEAI is present on our platform, the content is removed fromthe platform and reported to NCMEC, and we take enforcement action on the user account. Thisis detailed further in the Moderation and Enforcement sections of the Report. Upon knowledge ofany of the following activity, Snap will take enforcement action and report the user’s account toNCMEC: Snap works with NCMEC and other safety experts to learn about these types of harms and howthey may manifest themselves on our platform, and to report such harms to the proper authorities. Snap also has trusted flaggers to bring these and other types of harms to the attention of our trust and safety teams. There are also industry wide initiatives such as the TechCoalition’s Lantern Program which was launched on 10 November 2023.16 Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. 16 https://www.technologycoalition.org/newsroom/announcing-lantern. 47 CONFIDENTIAL DSA Mitigations Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, fundamental design decisions mean, for example, that Teens are less likely to come into contact with strangers. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our terms prohibit CSEAI and they are strictly enforced given the risk of severe harm. Our median turnaround time for child sexual exploitation reports in the second half of 2023 was 52 minutes. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove CSEAI. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend CSEAI i.e. there is no ‘CSEAI’ interest category. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, for example we have specific prevalence testing and monitoring moderation and enforcement data which we use to help detect and manage CSEAI-related risk. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to CSEAI/child safety. Codes and Crisis Protocols Yes, we cooperate with other providers through various groups e.g. EUIF, the 48 CONFIDENTIAL Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Technology Coalition, WeProtect Global Alliance. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), and how to get help in our Safety Center. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting. Our new parents site provides additional guidance for parents and carers on risks and support.17 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. We recognise there is growing concern regarding use of generative AI tools for CSEAI in general online. Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion Given the severity of the harm industry-wide, Snap still treats CSEAI as a Level 1 risk priority Inresponse to which it has put in place a range of mitigation measures. This includes in particular our proactive content moderation which is designed to detect and prevent CSEAI from appearing on each of Snapchat’s in-scope services – for example, our automated and human review onSpotlight. Our prevalence testing has continued to help us to improve this proactive contentmoderation. As a result, we’ve seen the prevalence of CSEAI on Snapchat fall to an extremely lowlevel. In addition, while we are alive to the risk, we have not identified any material issue withSnapchat’s in-scope services being used for the dissemination of generative CSEAI materials. 17 https://parents.snapchat.com. 49 CONFIDENTIAL We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for the dissemination of CSEAI. There is nochange in this conclusion from our 2023 Report. 4.1.2 Dissemination of illegal hate speech Public spaces displaying user generated content have the potential for the dissemination ofillegal hate speech. We recognise that, without mitigations, hate speech could conceivablyappear in any of Snapchat’s in-scope services displaying user generated content, from videosfeatured on Spotlight / Discover, to edits made to place labels on the Snap Map to changes to thenames and content of our Lenses and ads using hate speech or demeaning representations of aparticular culture, race or ethnicity. Likelihood Snap is sensitive to the issue of hate speech on internet platforms, as well as the damagingeffects hate speech can have on a community. Thankfully, hate speech is rarely found on thepublic surfaces of Snapchat. In our 2023 Report, we highlighted that our prevalence testing showed that hate speechaccounted for an extremely low percentage of total views of Snaps in Public Stories in August 2023 (see our Prevalence chapter). Recent assessments by European authorities have confirmed the low incidence of hate speech across Snapchat. A July 2023 report issued by ARCOMindicates that NGOs and other Trusted Flaggers submitted zero reports related to hateful contenton Snapchat over the course of the year.18 Snap Lenses are also not a popular medium for hatespeech, with an extremely low percentage of all reviewed Lenses falling within the Hate Speechcategory (all of which were rejected at submission). In this Report, we are pleased to confirm that our prevalence testing has shown that the PVP forHate Speech has continued to fall dramatically and is now at an extremely low level. The stepsSnap has taken to mitigate this harm mean that it is very unlikely that Snapchatters will encounterhate speech on Snapchat’s in-scope services, and Snap continues to place Hate Speech in the extremely low likelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that if illegal hate speech were to materialise on an online platform, the risk of harmwould fall within our ‘significant harm’ category. 18 ARCOM, ‘Lutte contre la diffusion de contenus haineux en ligne’, July 2023, url. 50 CONFIDENTIAL Overall potential risk prioritization As Snap continues to qualify hate speech as ‘significant’ in terms of severity but ‘lowest’ inlikelihood given the lowest relative prevalence on the platform, Snap considers hate speech a Level 3 risk prioritization. There is no change in this assessment from our 2023 Report. Snap's Mitigations Highlights Hate speech is strictly prohibited on Snapchat, and we continue to work with subject matterexperts, such as the FSM - Freiwillige Selbstkontrolle Multimedia Diensteanbieter e.V. in Germany,among others, to help identify hate speech, remove it from the platform, and take appropriateaction against users who post such content. Snap also works with law enforcement, whereappropriate, to take action against users who post illegal hate speech content. As explained in the Terms section of this Report, “hate speech” as defined in Snap’s Community Guidelines, includes both illegal and legal but harmful speech. As such, Snap’s definition of hate speech is more inclusive than most legal definitions of hate speech, because Snap wants totackle harmful (but) legal speech as well. It is possible, despite Snap’s terms and policies prohibiting such practices, as well as Snap’senforcement mechanisms, that malicious actors will find ways to circumvent Snap’s enforcementmechanisms and practices in order to post illegal hate speech, which could then appear onSnap’s public surfaces. As explained in our Moderation and Enforcement sections of our report, on our potentially high-reach surfaces, like Spotlight and Discover, we take a proactive approach to moderating anycontent that may violate our rules on hate speech. Our in-app reporting tool also allows users todirectly report hateful content or activities that support terrorism or violent extremism. Whenhateful content is reported, our teams will remove any violating content and users who engage inrepeated or egregious violations will have their account access locked. Lenses identified withhate speech were rejected when found during submission and disabled in Discover upon reviewif subsequently identified. As an additional measure, we encourage Snapchatters to block anyusers who make them feel unsafe or uncomfortable. Snap removes hate speech as soon as webecome aware of it, and will disable accounts dedicated to hate speech, hate symbols or groups,or the glorification of hate groups or members of a hate group. Our median turnaround time forhate speech reports in the second half of 2023 was 46 minutes. Our adaptation of Snapchat’s in-scope services to include moderation and enforcement tools andprocesses also encompasses service-specific adaptations to address illegal or violating contentsuch as illegal hate speech. 51 CONFIDENTIAL Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, our in-scope services have been adapted to include proactive moderation for illegal hate speech. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our terms prohibit illegal hate speech and they are strictly enforced. Our median turnaround time for illegal hate speech reports in the second half of 2023 was 46 minutes. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove illegal hate speech. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend illegal hate speech i.e. there are no interest categories relating to hate speech. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of Yes, for example we have specific prevalence testing and monitoring moderation and enforcement data which we use to help detect and manage illegal hate 52 CONFIDENTIAL their activities in particular as regards detection of systemic risk. speech. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to illegal hate speech, in particular Licra in France and the Department for Internet Services and Social Media. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various groups in relation to illegal hate speech. Snap remains a signatory of the EU Code of Conduct to counter illegal hate speech online and has worked hard to ensure Snap meets the requirements (including with respect to recent revision of that Code). Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on harms (see also Annex) and how to get help in our Safety Center. We make available robust reporting; and we provide guidance to parents on the web. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center, reporting and guidance. Our new parents site provides additional guidance for parents and carers on risks and support.19 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Yes, Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion Snap considers illegal hate speech a Level 3 risk prioritization. In response it has put in place arange of mitigation measures. These include in particular our alignment to the EU Hate Speech 19 https://parents.snapchat.com. 53 CONFIDENTIAL code of practice (including the recent revision) and our proactive content moderation which isdesigned to detect and prevent illegal hate speech from reaching a broad audience onSnapchat’s in-scope services. We monitor the prevalence of hate speech in general via our Prevalence Testing and external reporting which we publish in our Transparency Reports. As a result of the mitigation measures Snap has taken, hate speech continues to be an extremely lowprevalence risks. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for the dissemination of illegal hate speech.There is no change in this conclusion from our 2023 Report. 4.1.3 Dissemination of information related to the sale of prohibitedproducts or services The dissemination of information related to the sale of prohibited products or services (such asdangerous products, counterfeit products or illegally-traded animals) is a pervasive challenge fordigital platforms.On Snapchat, without mitigation, information related to the sale of prohibitedproducts or services could conceivably appear in any of Snapchat’s in-scope services displayinguser generated content, including information in videos featured on Spotlight / Discover andinformation about places to facilitate sales on Snap Map. It might also include ads promoting thesale of illegal goods, e.g. drugs or malicious content/malware. Likelihood Our testing shows that the prevalence of information related to the sale of prohibited products orservices on Snapchat continues to decrease year on year. According to our analysis, in July2024, content related to the sale of illegal goods is measured at an extremely low rate ofprevalence (PVP); content related to regulated (though not illegal) activities is measured at a extremely low rate,see our Prevalence section. This is a further, very significant decrease from that we reported for August 2023 in our 2023 Report. The likelihood of encountering suchviolating content on Snapchat’s in-scope services is now at its lowest level and falls within theextremely low likelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that if information related to the sale of prohibited products or services were tomaterialise on on an online platform, the risk of harm would fall within our: (i) ‘severe harm’ category where a credible threat to human life, safety, or well-being existed, inparticular the depiction or use of, or attempts at buying, selling, exchanging, or facilitating sales ofillegal lethal drugs 54 CONFIDENTIAL (ii) ‘substantial harm’ category where there are attempts to buy or sell weapons and depicting orbrandishing weapons in a threatening or violent context; and (iii) ‘significant harm’ category with respect to the dissemination of illegal or otherwise violatingcontent that relates to the use of online platforms for selling other illegal goods or services Overall potential risk prioritization Thankfully, as reported in our 2023 Report, this is still not a common issue on Snap. Ourprevalence testing revealed that communication around Illegal goods and activities now has onlya very small PVP rate (. However, due to the severity of some potential products and services(such as communication around dangerous or illicit drugs), prevalence is not the determinativefactor for Snap’s prioritization of this issue. Snap prioritizes severe harm and legal complianceover prevalence on the platform, and for this category has decided to deviate from the standardrisk framework. Snap would consider the overall risk of this type of content to be in the Level 1 category (due tothe level of severity) in cases where it concerns dangerous and illicit drugs, or any otherprohibited products or services that pose a threat to human life, safety, or well-being. Snapconsiders this issue a Level 2 overall potential risk prioritization in relation to weapons and a Level 3 potential overall risk prioritization in relation to other prohibited products and services.There is no change in this assessment since our 2023 Report. Snap's Mitigations Highlights Snap is sensitive to the issue of internet platforms being misused to advertise or sell prohibitedproducts or services. The steps Snap has taken to mitigate this harm have substantiallydiminished the likelihood that Snapchatters will find information related to prohibited products orservices on our platform. Snap’s Terms prohibit users from posting content that’s illegal in their jurisdiction or using Snap for any illegal activity. The Community Guidelines also prohibit promoting, facilitating, orparticipating in criminal activity, such as buying, selling, exchanging, or facilitating sales of illegalor regulated drugs, contraband (such as child sexual exploitation or abuse imagery), weapons, orcounterfeit goods or documents. They also prohibit promoting or facilitating any form ofexploitation, including sex trafficking, labor trafficking, or other human trafficking. Snap alsoprohibits the illegal promotion of regulated goods or industries, including unauthorized promotionof gambling, tobacco or vape products, and alcohol. Across Snapchat, we offer a number of resources to users to raise awareness on safety topics and protect them. As explained in the Transparency section of the Report, one of the examples of this is our in-app tool, Heads Up. This surfaces educational content from experts to Snapchattersif they try to search for drug-related content. Our expert partners include the Centers for Disease 55 CONFIDENTIAL Control and Prevention (CDC), the Substance Abuse and Mental Health Services Administration(SAMHSA), Community Anti-Drug Coalitions of America (CADCA), Shatterproof, Truth Initiative,and the SAFE Project. It is possible, despite Snap’s terms and policies prohibiting such practices and offering support,as well as Snap’s enforcement mechanisms, that malicious actors will find ways to circumventSnap’s enforcement mechanisms and practices in order to post information related to the sale ofprohibited products or services, which could then appear on Snap’s public surfaces. As explained in the Moderation section of the Report, we have proactive and reactive moderation processes in place to detect and moderate content relating to the sale of illegal goods andservices, and we have aggressively focused on enforcement of severe and serious harms. Forexample, during the second half of 2023, we enforced against more than 365,000 pieces ofcontent and more than 240,000 accounts relating to drug content, based on both proactivedetection of drug sales content and reports in-app and through the support site. Moreover, ourenforcement of accounts for violating our Community Guidelines relating to weapons hascontinued to double year on year during the same period. We also use Prevalence Testing to continuously improve our moderation. We continue to use violating drug Snaps discovered from PVP sampling, which we consider a Level 1 potential overallrisk prioritization for illegal goods and services, to help train our proactive machine learningdetection models. As a result of this and other efforts since our 2023 Report, proactive detectionand enforcement volumes for violating drug Snaps has increased. As explained in the Enforcement section of our Report, Snap complies with relevant legal requirements to remove content about the sale of illegal goods and services, and takesappropriate action against egregious or repeat violators. Snap works with law enforcement,safety organizations, and subject matter experts to continue to educate ourselves and ourcommunity, and to take appropriate action where these threats may arise on our platform. When we identify violators engaging in the attempted buying, selling, exchanging, or facilitatingsales of dangerous and illicit drugs, we disable their accounts and, in some instances, refer theconduct to law enforcement. For less severe harms, a user will be warned and their contentremoved. Repeat violations will result in violators’ accounts being disabled. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of the 56 CONFIDENTIAL specific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, design decisions, including adding proactive moderation to Snapchat’s in-scope services, make it difficult for the sale of prohibited products or services to reach a large audience. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, terms prohibit the sale of prohibited products or services and they are strictly enforced with the most serious consequences. Our median turnaround time for violating content relating to illegal and other regulated goods were all less than 60 mins in the second half of 2023 (32 minutes for regulated goods, 35 mins for weapons and 55 mins for drugs). Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent the sale of prohibited products or services. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend content concerning the sale of prohibited products or services content i.e. there is no interest category for this content. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, for example we have specific prevalence testing and monitoring moderation and enforcement data which we use to help detect and manage information related to the sales of prohibited products and services. 57 CONFIDENTIAL Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to the sale of prohibited products or services, in particular the Danish Safety Technology Authority. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups and share signals, especially in relation to drug dealers with the EU Internet Forum which has recently expanded its work to tackle drug sales online. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our in-app tool, Heads Up, surfaces educational content from experts and we try to flag this resource to Snapchatters if they search for drug-related content. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers, Family Center, reporting, and guidance. Our new parents site provides additional guidance for parents and carers on risks and support.20 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. General content authenticity measures Conclusion Despite the continued very low prevalence, we still consider the overall risk of the disseminationof the sale of dangerous or illicit drugs, or any other prohibited products or services that pose athreat to human life, safety, or well-being, to be in the Level 1 category due to the level of severity. 20 https://parents.snapchat.com. 58 CONFIDENTIAL Snap continues to consider that the sale of weapons poses a Level 2 overall potential risk, and aLevel 3 potential overall risk in relation to other prohibited products and services. Snap continuesto take steps to mitigate these harms, which has further diminished the likelihood thatSnapchatters will find information related to prohibited products or services on Snapchat’sin-scope services. Snap continues to invest significant resources to further combat these harms,and are still looking to achieve further reductions in the likelihood of this risk where possible. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for dissemination of information related tothe sale of prohibited products or services. There is no change in this conclusion since our2023 Report. 4.1.4 Dissemination of terrorist content As noted in our 2023 Report, online influences have been depicted as major drivers for thepropagation and adoption of extremist ideologies, which often contain an element of collectivegrievance, and subsequent acts of violence.21 It is conceivable that, without mitigations, badactors could disseminate terrorist content on Snapchat, as with any other online platform. Thiscould include, in particular, terrorist content appearing in videos featured on Spotlight / Discoverand extremist content and individuals promoted via Public Profiles. Likelihood According to our testing, terrorist content is measured to have an extremely low prevalence onour platform – see our Prevalence chapter. This is a slight increase from the prevalence rate weobserved in our 2023 Report (although still extremely low). In the second half of 2023, forexample, we removed 320 accounts in the EU for violations of our policy prohibiting terrorist and violent extremist content, as recorded in our Transparency Reports. This is also a slight increase compared with 23 for the second half of 2022, which we observed in our 2023 Report. Theseincreases are consistent with the slight increase in terrorism related content we observedfollowing the attacks in Israel on 7 October 2023 and the subsequent conflict in Gaza - Israel. As reported in our 2023 Report, we have previously sought independent analysis via third partyintelligence vendors that track extremist activity online who have verified that Snapchat does notfall into the top 100 communications platforms used by extremist groups to communicate. Morerecently, in response to the various events influencing violent extremism and terrorism onlinethrough 2023, about 349 removal orders were sent by 6 EU member States' competent 21 J.F. Bender and J. Kenyon, Terrorism and the internet: How dangerous is online radicalization?, Front.Psychol., 13 October 2022, url. 59 CONFIDENTIAL authorities to 13 online platforms under the Regulation on dissemination of terrorism contentonline;22 none were directed at Snap.23 These data indicate that, while there has been a slight increase in the likelihood of encounteringterrorist content on Snapchat, it remains extremely low and it still falls within the lowestlikelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that, if information related to terrorist content were to materialise on an online platform,the risk of harm would fall within our ‘severe harm’ category due to the high threat to human life,safety, or well-being. Overall potential risk prioritization Due to the very low prevalence of extremist content on Snapchat, the overall risk would normallybe assessed to be Level 3. However, Snap has decided to deviate from the standard riskframework, and has placed terrorist content within our Level 2 overall potential risk prioritizationcategory (and Snap will always consider the overall risk to be Level 1 risk prioritization wherethere is an immediate risk to human life, safety, or well-being) There is no change in thisassessment from our 2023 Report. Snap's Mitigations Highlights Snap is sensitive to the issue of dissemination of terrorist content on internet platforms andservices. The steps Snap has taken to mitigate this risk have substantially diminished thelikelihood that Snapchatters will encounter terrorist content. In addition, unlike many of our peers,Snap does not offer an open news feed where unvetted publishers or individuals have anopportunity to broadcast terrorist content, nor does Snapchat offer a ‘reshare’ functionality thatwould encourage virality, and does not allow user-generated content to gain wide viewershipwithout going through human review. Snap’s Terms and Community Guidelines expressly prohibit terrorist organizations, violent extremists, and hate groups from using our platform. We consult the expertise and work of civilrights organizations, human rights experts, law enforcement agencies, NGOs, and safetyadvocates to help enforce these Guidelines. Such expert knowledge comes from sources such asthe Anti-Defamation League, the Southern Poverty Law Center, the Election Integrity Partnership,the Atlantic Council, the Stanford Cyber Policy Center, the members of Snap’s Safety Advisory 23 Violent Extremism and Terrorism Online in 2023: The Year in Review, April 2024, url. 22 Regulation (EU) 2021/784. 60 CONFIDENTIAL Board, and individual domain experts (including a former Ambassador to the UN Human RightsCouncil, leading digital rights scholars and advocates, former regulators and policymakers, andgeopolitical experts). We are constantly learning, and will calibrate wherever necessary to ensurethat our products and policies function to keep Snapchatters safe. Our prohibitions against Terrorism and Violent Extremism extend to all forms of content thatpromotes terrorism or other violent, criminal acts committed by individuals or groups to furtherideological goals. These rules also prohibit any content that promotes or supports foreignterrorist organizations or extremist hate groups––as designated by credible, third-partyexperts––as well as recruitment for such organizations or violent extremist activities. It is possible, despite Snap’s terms and policies prohibiting such practices, as well as Snap’senforcement mechanisms, that malicious actors will find ways to circumvent Snap’s enforcementmechanisms and practices in order to post terrorist content, which could then appear on Snap’spublic surfaces. As explained in our Moderation section, on our high-reach surfaces, like Spotlight and Discover,we take a proactive approach to moderating any content that may violate these rules. Our in-app reporting tool allows users to directly report hateful content or activities that supportterrorism or violent extremism. Snap removes such content, disables accounts, and cooperateswith law enforcement as such issues may arise; see our Enforcement section for moreinformation. Users engaged in terrorist activities or violent extremism will lose account privileges.Accounts we discover engaging in the following activity will immediately be disabled and whereappropriate, reported to law enforcement. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat is not an attractive platform for terrorist content because it is difficult to reach a large audience on Snapchat, and Snap proactively moderates Snapchat’s 61 CONFIDENTIAL in-scope services that provide an opportunity to reach a larger audience. As a result, we experience very few instances of terrorist content on Snapchat. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our terms prohibit terrorist content and they are strictly enforced with the most serious consequences. Our median turnaround time for Terrorism \& Violent Extremism reports in the second half of 2023 was 45 minutes. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove terrorist content and accounts. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend terrorism content i.e. there is no ‘terrorism’ interest category. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, for example we have specific prevalence testing and transparency reporting which we use to help detect and manage terrorist content. We have also sought independent analysis via third party intelligence vendors that track extremist activity online who have verified that Snapchat does not fall into the top 100 communications platforms used by extremist groups to communicate. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. No, we don’t have a specific trusted flagger group we currently work with on terrorism content in the European Union. This is due to the low prevalence of terrorist content on Snap. 62 CONFIDENTIAL Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups, including the EU Internet Forum (EUIF), that consider terrorist content. Note, due to the low prevalence of terrorist content on Snap, we do not participate in the primary multi stakeholder organization: The Global Internet Forum to Counter Terrorism (GIFCT). Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on harms (see also Annex) and how to get help in our Safety Center. We make available robust reporting; and we provide guidance to parents on the web. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center, reporting and guidance. Our new parents site provides additional guidance for parents and carers on risks and support.24 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Yes, Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion Despite the very low prevalence, we consider the potential risk of the dissemination of terroristcontent to be in the Level 2 risk prioritization category due to the level of severity. Snapchat’sdesign and its proactive detection measures make Snap a very unpopular place for thedissemination of terrorist content. The prevalence and enforcement rates for terrorist contenthave experienced slight increases since our 2023 Report but there continues to be a very lowlikelihood of users being exposed to this illegal content. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for dissemination of terrorist content.However, due to the slight increase in prevalence and enforcement rates since our 2023 24 https://parents.snapchat.com. 63 CONFIDENTIAL Report, consistent with a slight risk in terrorism related activity we observed following the 7October attacks in Israel, we continue to carefully monitor this risk category to ensureprevalence remains very low. 4.1.5 Dissemination of content that infringes on intellectual propertyrights All platforms that allow users to upload and share media have the potential for those users tochoose to upload material that they do not have the right to share (for example, clips from films),or that users may create and share original material that infringes on another party’s intellectualproperty (for example, a Lens using a copyrighted character). Without mitigations, such materialcould conceivably appear on Snap's public surfaces including in particular videos on Spotlightand Discover. Likelihood Snapchat’s platform architecture does not favor the mass distribution of unauthorizedcopyrighted content. Snapchat does not have a live-streaming feature. A typical “Snap” is 10seconds or less and expires in 24 hours. Content creation and consumption on Snapchat favorsvery short, original content and in-the-moment communication between friends; other platformsare more attractive to those seeking to flout intellectual property law. Snap maintains a public Transparency Report which includes data on enforcement actions related to intellectual property infringement. In H1 of 2023: ● We received 1,159 copyright notices; 49% of those requests led to the removal ofsome content. This compares with 558 copyright notices - 78% of those requestsleading to the removal of some content - in H1 2022. ● We received 224 trademark notices; 28% of those requests led to the removal ofsome content. This compares with 96 trademark notices - 29% of those requestsleading to the removal of some content - in H1 2022. In H2 of 2023: ● We received 1297 copyright notices; 57% of those requests led to the removal ofsome content. This compares with 905 copyright notices - 73% of those requestsleading to the removal of some content - in H2 2022. ● We received 203 trademark notices; 23% of those requests led to the removal ofsome content. This compares with 172 trademark notices - 13% of those requestsled to the removal of some content - in H2 2022. 64 CONFIDENTIAL This data continues to show slight overall increases in reports of intellectual property issues yearon year but a consistently low prevalence in absolute terms. As a result, Snap continues toconsider the likelihood of encountering content that infringes intellectual property on Snapchat iswithin the extremely low likelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that if information infringes intellectual property rights were to materialise on an onlineplatform, the risk of harm would fall within our ‘significant harm’ category Overall potential risk prioritization We consider the dissemination of content that infringes on intellectual property rights is a Level 3overall potential risk on Snapchat. We take reports seriously and the reported infringement ofintellectual property often leads to content removal or, in some cases, the deletion of the user’saccount. There is no change in this assessment since our 2023 Report. Snap's Mitigations Highlights Snapchat respects the intellectual property of others, and expects our users to do the same. As we explain in the Terms section of the Report, Snap’s Terms of Service clearly prohibit the use of Snap’s services to infringe on someone else’s intellectual property rights. The Enforcement section of the Report states that if someone believes that any content onSnapchat infringes their intellectual property (IP), they can let us know via our reporting menu or online forms for Copyright Infringement or Trademark Infringement. Snap honors copyright laws, including the Digital Millennium Copyright Act and European Copyright Directive, and takesreasonable steps to expeditiously remove from our Services any infringing material that webecome aware of. If Snap becomes aware that a user has repeatedly infringed copyrights, we willtake reasonable steps within our power to suspend or terminate the violator’s account. Snap respects the doctrine of “fair use,” (where applicable) i.e., that there are certaincircumstances (such as news reporting, social commentary on issues of public interest, criticism,parody, or education) where copyrighted material could be distributed without permission from,or payment to, the copyright holder. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigation 65 CONFIDENTIAL category applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, content on Snapchat is typically short in nature, the average Snap is 10 seconds, and reporting tools help with the detection of IP infringing material. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, terms prohibit IP infringements and they are strictly enforced. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific reactive moderation procedures to expeditiously remove content that infringes intellectual property rights. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Our algorithmic systems do not knowingly recommend content that infringes intellectual property rights, i.e. there are no interest categories relating to specific intellectual property. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we have a notice procedure to flag and enable us to respond to intellectual property infringements. Trusted Flaggers General trusted flagger measures. 66 CONFIDENTIAL Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Not applicable. We respond to reports of infringement on an individual basis. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we warn users not to publish content that infringes on intellectual property rights and we have an easily accessible reporting tool. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. No specific measures relating to the protection of minors for this risk. Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Yes, Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion We consider the overall risk of the dissemination of IP infringing content to be significant. Snaphas taken steps to mitigate these harms, which has substantially diminished the likelihood thatSnapchatters will encounter IP infringing material. These mitigations include product and designmeasures like short content retention periods, some proactive moderation, andnotice-and-takedown procedures. 67 CONFIDENTIAL We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for dissemination of content that infringesintellectual property rights. There is no change in this conclusion since our 2023 Report. 4.1.6 Dissemination of adult sexual content Estimates as to the volume of adult sexual content on the Internet vary, but some historicalstudies have considered that around 4% of websites, 13% of web searches and 20% of mobilesearches were related to adult sexual content25. As such it is conceivable that, withoutmitigations, this content could also appear on any of Snapchat’s in-scope services including inparticular videos on Spotlight, Discover, promoted on Public Profiles, features as part of ourLenses or as places on the Snap Map and be the subject of advertisements via Snap Ads. Likelihood As shown in our Prevalence Testing, prevalence of adult sexual content continues to be on adownward trajectory confirming that ongoing detection and enforcement enhancements havebeen having the desired impact. Although adult sexual content is still our most prevalent illegal or violating content compared withthe other risks on Snapchat’s in-scope services, we are very pleased to have been able toachieve significant further reductions in prevalence since our 2023 Report. The prevalence issignificantly lower than the prevalence of adult content on the Internet in general. Further, as reported in our Transparency Reports, Sexually Explicit Content has seen a steady reduction in the % of the total content enforced by Snap: Period % of the Total Content Enforced by Snap 1H 2022 76.6% 2H 2022 67.9% 1H 2023 52.6% 2H 2023 42.1% As a result of these substantial reductions in both PVP and proportion of enforced content, in thisreport we are now placing adult sexual content in our Very Low likelihood category. This is a 25 Ogas, O. and S. Gaddam (2012), Boston University, A Billion Wicked Thoughts: What the Internet Tells UsAbout Sex and Relationships; and Google Inc, Columbia University and Carnegie Mellon University, ALarge Scale Study of Wireless Search Behaviour, 2005. 68 CONFIDENTIAL decrease from our 2023 Report when adult sexual content was placed in the low likelihoodcategory. We continue not to identify any significant volumes of adult sexual crimes and so westill place this in our extremely low likelihood category. Severity Snap has assessed information published by governments and third party sources and considersthat the severity of this risk varies depending on the nature of the content as follows: ● If information relating to adult sexual offences were to materialise on an online platform,the risk of harm would fall within our ‘serious harm’ category due to the significant threatto human life and well-being, abusing people’s fundamental rights and dignity andinvolving the criminal exploitation of vulnerable people. ● If information relating to sexually explicit content or depictions of nudity were tomaterialise on an online platform, the risk of harm would fall within our ‘significant harm’category. This content, while significant, does not pose the same severity of risk as adultsexual offences. Overall potential risk prioritization The overall potential risk of this adult sexual content depends primarily on severity of the issue.Overall, we consider the dissemination of sexual crimes and offenses on Snapchat’s in-scopeservices to be a Level 2 risk on Snapchat. We consider the dissemination of sexually explicitcontent or depictions of nudity to be a Level 3 potential overall risk prioritization. There is nochange in this assessment since our 2023 Report. Snap's Mitigations Highlights Snap’s Terms prohibit promoting, distributing, or sharing pornographic content, as well ascommercial activities that relate to pornography or sexual interactions (whether online or offline).Breastfeeding and other depictions of adult nudity in non-sexual contexts are generallypermitted. As this can be a challenging area, we make available additional guidance on sexual conduct and content that violates our Community Guidelines here. As explained in the Moderation and Enforcement sections of the Report, we have proactive andreactive moderation processes in place to detect and moderate adult sexual content. Our in-appreporting tool allows users to directly report adult sexual content which our teams will thenremove if confirmed as violating. 69 CONFIDENTIAL Our Prevalence Testing has continued to have a very significant impact in reducing the extent towhich adult sexual content is present on Snapchat to low levels. We have seen further substantialdecreases in the prevalence of adult sexual content since our 2023 Report confirming thatongoing detection and enforcement enhancements have been having the desired impact. Thishas resulted in a further reduction in prevalence. In addition to measures taken against sexual crimes and sexually explicit content (and nudity),Snap also takes steps to limit the prevalence and recommendation of sexually suggestivecontent. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, our online platforms have been designed to limit the prevalence of sexually suggestive content. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our terms prohibit adult sexual content and they are strictly enforced. Our median turnaround time for adult sexual content reports in the second half of 2023 was 7 minutes. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove adult sexual content. Algorithmic Systems Yes, our algorithmic systems do not knowingly recommend adult sexual content 70 CONFIDENTIAL Testing and adapting their algorithmic systems, including their recommender systems. i.e. there is no ‘adult sexual content’ interest category. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, for example we have specific prevalence testing and transparency reporting which we use to help detect and manage adult sexual content. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to Non-Consensual Intimate Image Abuse (NCII), notably Stop Fisha in France. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups, including in particular the EU Internet Forum (EUIF) which has expanded its remit to also tackle the trafficking of human beings (which is often driven by sexual crimes or pornography). Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on harms (see also Annex) and how to get help in our Safety Center. We make available robust reporting mechanisms; and we provide guidance to parents on the web. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center, reporting and guidance. Our new parents site provides additional guidance for parents and carers on risks and support.26 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or Yes, Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are 26 https://parents.snapchat.com. 71 CONFIDENTIAL truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. disseminated on Snapchat’s inscope services. Conclusion Adult sexual content is our highest prevalence, largest enforced content category issue on Snapchat’s in-scope services, and represents a Level 3 overall potential risk. We also treat sexualcrimes as a Level 2 risk given the risk of serious harm, despite Snap considering this to fall withinthe extremely low likelihood category. We have continued to dedicate substantial resources andtaken significant steps, including further improvements to our proactive detection mechanisms.This has resulted in further substantial decreases in the prevalence rates for adult sexual content. This significant proactive detection work was undertaken as a direct response to us detecting anuptick in the prevalence of adult sexual content on Snapchat, which demonstrates theeffectiveness of our risk detection and management framework and procedures. We continue towork towards further reductions in the prevalence and enforced content percentages for adultsexual content. With these adjustments, and the other specific mitigations listed above, we have concludedthat Snapchat’s in-scope services have reasonable, proportionate and effective mitigationmeasures for dissemination of adult sexual content. We will continue to carefully monitor thisrisk category as we look to achieve further reductions in adult sexual content prevalence. 4.1.7 Dissemination of content regarding harassment \& bullying Unfortunately, harassment and bullying that have always been a persistent problem in schoolsand workplaces have crossed over to the online environment with the growth of the Internet.Without mitigations, this content could conceivably appear on Snapchat’s in-scope services.Likelihood In our 2023 Report, we concluded that we would continue to monitor the prevalence ofharassment \& bullying content as we had seen a rise in prevalence and reporting. In particular, as we observed a slight increase in PVP for harassment (which includes bullying). At that timeharassment was in the very low risk category observed on Snapchat’s in-scope services and wewished to determine whether the rise in prevalence and reporting was a temporary issue due tothe design of new reporting options, or if additional measures need to be put in place. We have since made reporting flow enhancements that simplified Snapchatters ability to reportharassment \& bullying content and have also expanded the scope of reporting. We are pleased 72 CONFIDENTIAL to say that, while harassment remains within the very low category, since our 2023 Report wehave observed a significant reduction in PVP. Our Transparency Report shows that harassment is an issue that leads to a significant volume of content and account enforcements : ● In the first half of 2022, we received approximately 150,994 reports related to harassmentin the EU, of which roughly 31% resulted in some enforcement action. In the second half of2022, we have received significantly more reports, 566,708, relating to harassment. Thissignificant increase was due to the introduction of new reporting options. ● We also noted that in the second half of 2022, only 13% resulted in some enforcementaction. This means that while the new reporting tools surfaced more reports, they appearto have reduced the overall quality of those reports. As a result we were investigating thecause of this but at the time of our 2023 Report we believed this was due to “harassmentand bullying” being the first reporting option in the reporting menu. In addition, we notedthat we received a significant number of reports of “harassment” in bad faith i.e. where anindividual reports public, amplified content because they don’t like it - for example a usermay dislike that one prominent person is criticizing the actions or views of anotherprominent person, and report their content as “harassment.” In other instances, we mayreceive a report alleging “harassment” where there is not sufficient information to takeenforcement action. This accounts for significant differences between reporting andenforcement rates in general, and in particular for the new reporting tools whereharassment and bullying may be selected as the first option regardless of whether thereport concerns such behavior. The above-discussed trends continued through the second half of 2023, in which we receivedeven more reports related to harassment in the EU: 1,371,668 reports, with only 16% of thoseresulting in some enforcement action. We believe that the sharp increase in the number of totalreports continued to be due to the wider use and rollout of new reporting options, as well as theissues mentioned in the preceding paragraph. Additionally, we believe that the Israel-Hamasconflict, and resulting anti-Israel and anti-Hamas sentiment, may have had an impact as well. Weadjusted our reporting menu to place “harassment and bullying” as the second (and not first)option on the reporting menu (second to “I just don’t like it”); however, this change does not seemto have had much of an impact on report volumes. While we continue to see significant increases in the content reported and enforced asharassment \& bullying content, as in our 2023 Report, we would flag that the bulk of trueharassment occurs in non-public surfaces which are out of scope of our DSA risk assessment. When we consider Snapchat’s in-scope services specifically, i.e. the public parts of Snapchatwhich fall within the scope of the risk assessment obligations under Article 34, the likelihood ofthese public spaces being used for the dissemination of bullying \& harassment content is evenlower. In 2023, we rejected 1903 Snaps on Spotlight and 3065 Snaps on Discover. We have seen 73 CONFIDENTIAL a significant fall in the prevalence of harassment \& bullying content from our testing of thein-scope services of Snapchat. As a result, in respect of Snapchat’s online platforms within thescope of our DSA risk assessment only, we consider harassment \& bullying to still fall within our Very Low likelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that if information relating to adult sexual offences were to materialise on an onlineplatform: (i) where harassment and bullying involve both (1) harms that risk significant damage to thephysical or emotional well-being of Snapchatters, and (2) the imminent, credible risk of severeharm, including threats to human life, safety, and well-being, it would fall within our ‘severe harm’category; and (ii) in general, would fall within our ‘serious harm’ category. Overall potential risk prioritization In general, we have assessed the dissemination of content regarding harassment \& bullying tofall within our Level 2 potential risk prioritization categoryHowever, all situations where there isrisk of significant damage and an imminent risk of severe harm, are considered to fall within ouroverall Level 1 potential risk prioritization category. There is no change in this assessment sinceour 2023 Report. Snap’s Mitigations Highlights Snap’s policies prohibit a range of content or behavior that harasses individuals, including: (1)harassment and bullying in general, (2) behaviour that constitutes or promotes sexualharassment; or (3) behaviour that constitutes non-consensual intimate content (i.e., productionand/or distribution). When we consider whether to allow content for algorithmic recommendations, we apply additional rules. In practice, where we algorithmically recommend content on our online platforms, we takeproactive measures to stop the dissemination of content that includes harassment \& bullying. Weuse a mix of automation (such as abusive language detection, image recognition models, andaccount history) and human review to enforce our guidelines as explained in the Moderationsection of this Report. In our Enforcement section, we also explain the significant resourcesdevoted to preventing the dissemination of content that includes harassment \& bullying. Anycontent anywhere on Snapchat can be reported in-app or on our web site, and “harassment” is 74 CONFIDENTIAL one of the reporting reasons offered, and as reported above, this includes new reporting optionsfor user profiles. Importantly, our Transparency Report shows that the median turnaround time for a harassment report is 7 minutes. If content that constitutes harassment and bullying is reported to us, werespond very swiftly with appropriate action. Additionally, when we learn of content suggesting that there is an emergency situation involvingimminent danger of death or serious bodily injury involving any person, we will proactivelyescalate the report to law enforcement. We have established channels for referring such contentto the FBI in the U.S. and Interpol in the rest of the world. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, fundamental design decisions mean that content constituting harassment and bullying can be easily reported. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our terms prohibit harassment \& bullying. This is explained to users clearly in our Harassment \& Bullying explainer with guidance on how we apply this policy. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, we have specific moderation procedures to prevent and remove harassment \& bullying content. 75 CONFIDENTIAL Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend harassment \& bullying i.e. there is no ‘bullying and harassment’ interest category. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, for example we have specific prevalence testing and monitoring moderation and enforcement data which we use to help detect and manage harassment \& bullying risk. Our Safety Advisor Board also has several anti-bully experts which we call on for independent review and expertise. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with many trusted flaggers in the EU in relation to child safety, including for example E-Enfance in France. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. We are not working with other providers on harassment and bullying specifically. However, we work with several groups in relation to child protection in general, including in relation to the new EU AAD Code. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. We provide local resources related to bullying and harassment. In France for example we direct users to E-Enfance. The new national number against digital violence, free for children and adolescents facing problems related to their digital use-- 100% anonymous, free and confidential. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting; and we provide guidance to parents and teens, including the safety measures and resources highlighted in the Transparency mitigation section above, such as the Harassment and Bullying explainers. 76 CONFIDENTIAL Our new parents site provides additional guidance for parents and carers on risks and support.27 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. General content authenticity measures. Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. This includes applying an AI sparkle icon in specific situations, such as our Bitmoji Backgrounds. We continue to assess whether to include such an icon on a case by case basis, considering whether generated images are photorealistic. Conclusion Harassment and bullying is the second most prevalent issue faced by Snapchat. However, it isstill a low likelihood in absolute terms. Where there is a risk of severe harm, we consider bullyingand harassment has a Level 1 overall potential risk. In general, we consider the dissemination ofcontent on Snapchat’s in-scope services that includes harassment \& bullying is a Level 2potential risk on Snapchat. In practice, we have taken significant measures to prevent harassmentand bullying, including clear guidance on our rules and how we enforce them, easy to accessreporting tools and very rapid response times to address violating content. New reporting optionshave continued to result in a significant rise in reports combined with a fall in the enforcementrate. We are still investigating the reasons for this but believe these continue to relate to the outof scope services on Snapchat. Our prevalence testing has shown significant further reductions inthe PVP rate for harassment and bullying for Snapchat’s in-scope services. As a result, we have concluded that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for dissemination of harassment and bullyingcontent. While we have seen a significant fall in the prevalence of harassment \& bullyingcontent from our testing of the in-scope services of Snapchat, we will continue to carefullymonitor this risk category as we look to achieve further reductions. 4.1.8 Dissemination of content that glorifies self-harm, including suicide The risk of young people encountering content that promotes glorifies self-harm, including thepromotion of self-injury, suicide, eating disorders, body image dissatisfaction and distorted values 27 https://parents.snapchat.com. 77 CONFIDENTIAL and attitudes online and on social media in general has been identified in several studies.Without mitigation, this content could conceivably appear on Snapchat’s in-scope services. Likelihood In our 2023 Report, we concluded that on this in-scope services of Snapchat, the disseminationof self-harm and suicide content fell within our Extremely Low likehood category due to the lowrates prevalance and reporting. Since our 2023 Report: ● Snap has assessed information published by governments and other third party sourcesrelating to the relative likelihood of self-harm content on online platforms, includingSnapchat. ● Our Transparency Report continues to show that “self harm \& suicide” is an issue that still leads to a moderate volume of content and account enforcements. In the second half of2023, we received 188,124 reports related to self-harm and suicide (up from 129,785reports in the second half of 2022), enforcing against approximately 32,841 pieces ofcontent (up from 20,054 in 2H 2022) and 28,207 accounts (up from 18,311 in 2H 2022).However, as we noted in the 2023 Report, those figures relate to Snapchat in general.When we consider Snapchat’s in-scope services specifically, i.e. the public parts ofSnapchat which fall within the scope of the risk assessment obligations under Article 34,the likelihood of these public spaces being used for the dissemination of content thatglorifies self-harm is very low. In 2023, we rejected 4248 Snaps on Spotlight and 110Snaps on Discover. ● Our prevalence testing continue to show very low prevalence rates for self-harm andsuicide content. This continues the trend we highlight in our 2023 Report where weobserved a significant decrease to negligible rates in August 2023. Very few Lenses haveever been found to include self-harm content and all of these were rejected beforepublication. As a result, we consider the risk of dissemination of content glorifying self-harm to fall within our extremely low likelihood category for Snapchat’s in-scope services. Severity Snap has assessed information published by governments and other third party sources andconsiders that if information relating to self-harm were to materialise on an online platform, weconsider the severity of harm risked from such content (including content relating to self-injury,suicide or eating disorders) to fall within our “serious harm” category, Where the disseminationof content that indicates an imminent, credible risk of severe harm, including threats to humanlife, safety, and well-being, we consider the severity of harm risked to fall within our severe harm 78 CONFIDENTIAL category (as explained in our severe harm explainer) In practice, we devote enforcement resources to preventing the dissemination of content that glorifies self-harm, including thepromotion of self-injury, suicide or eating disorders. Overall potential risk prioritization Although the prevalence of content that glorifies self-harm on Snapchat’s in-scope services isconsidered to be at the lowest level of all our risks, due to the potential for severe and seriousharms to be caused, we have chosen to elevate the risk prioritization for these risks. Snap willalways consider the dissemination of content that indicates an imminent, credible risk of severeharm, including threats to human life, safety, and well-being, as Level 1 overall potential risk prioritization (as explained in our severe harm explainer), and we devote significant resources to combatting this type of harm. Other content relating to self-harm (including content relating to self-injury, suicide or eatingdisorders) are also classified as a Level 2 risk prioritization overall. As described in our riskmethodology section, we assess overall potential risk on a case by case basis and Snap reservesthe option to deviate from the overall potential risk prioritization matrix we use as a guide. This isone of the cases where we have chosen to deviate. There is no change in this assessment sinceour 2023 Report. Snap's Mitigations Highlights Snap’s Terms prohibit the dissemination of content that promotes self-harm and suicide. On Snapchat’s in-scope services where we algorithmically recommend content, we takeproactive measures to stop the dissemination of content that glorifies self-harm, including thepromotion of self-injury, suicide or eating disorders. We allow some discussion (such as news orpublic issue commentary) of self-harm, suicide, or eating disorders, when the discussion is notglorifying such behavior. Even so, we mark this content as “sensitive” internally and adjust ouralgorithmic systems to limit recommendations of this kind of content. As described in our Moderation and Enforcement sections of this Report, we use a mix ofautomation (such as abusive language detection, image recognition models, and account history)and human review to enforce our guidelines. Since our 2023 Report, we have continued to makeimprovements to our proactive detection tools, including with respect to self harm and suicidecontent, which has likely contributed to the substantial reduction in prevelance of public Storieson Snapchat’s inscope services. 79 CONFIDENTIAL Any content anywhere on Snapchat can be reported in-app or on our web site, and “self-harm and suicide” is one of the reporting reasons offered. Our Transparency Report shows that the median turnaround time for a “self-harm and suicide” report is 44 minutes. When we learn of content suggesting that there is an emergency situation involving imminentdanger of death or serious bodily injury involving any person, we will proactively escalate thereport to law enforcement. We have established channels for referring such content to the FBI inthe U.S. and Interpol in the rest of the world. We also work with third-party mental health groups to surface supportive interventions in-app. Auser who searches for certain terms related to self-harm or suicide may be routed to suicidehelplines in their region. For example, Snap has established a self harm flow for Lenses, whichincludes escalation to the Trust \& Safety team, sending help resources and escalation to the LawEnforcement Operations team. Lenses that are rejected, although few in number, include helpresources within the rejection reason. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive moderation for content that promotes self-harm. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our terms prohibit content that promotes self-harm and they are strictly enforced. Our Transparency Report shows that the median turnaround time for a “self-harm and suicide” report is 44 minutes. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in Yes, we have specific proactive and reactive moderation procedures to prevent and remove content that promotes self-harm. Snap includes help resources within rejection reasons, for example, in the Lenses submission flow. 80 CONFIDENTIAL particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend content glorifying self-harm i.e. there is no ‘glorifying self-harm’ interest category. We mark non-glorifying discussion as sensitive and limit the volume of recommendations. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, for example we have specific prevalence testing and monitoring moderation and enforcement data which we use to help detect and manage self-harm and suicide. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to content that promotes self-harm. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. In Snapchat we provide a number of tools to users. For example, if a user searches for suicide related terms we will surface our Here For You tool. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and 81 CONFIDENTIAL carers on risks and support.28 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. General content authenticity measures. Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion Content that glorifies self-harm content is categorized within the Extremely Low likelihoodcategory for Snapchat’s in-scope services. However, this content falls within our ‘serious harm’category and as a result we have decided to categorize it as a Level 2 overall potential risk, eventhough our risk matrix would suggest a lower category. We always treat content relating tosuicide and other situations involving imminent, credible risk of harm as a Level 1 overall potentialrisk. In response, we have significant dedicated mitigation measures, including clear prohibitions,guidance, proactive and reactive moderation, reporting tools, sensitive content recommendationlimits and cooperation with trusted flaggers. We respond rapidly to reports of self-harm, with amedian turnaround time of less than one hour. Our prevalence rates for self-harm content onSnap’s in-scope services continue to decline and are at very low levels. As a result, we have concluded that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for dissemination of content glorifyingself-harm (including the promotion of self-injury, suicide or eating disorders). Snap monitorsthis category to confirm whether further mitigating measures might be required. There is nochange in this conclusion since our 2023 Report. 4.1.9 Dissemination of content relating to violent or dangerous behavior Without mitigations, content encouraging or engaging in violent or dangerous behavior couldconceivably appear on Snapchat’s in-scope services. 28 https://parents.snapchat.com. 82 CONFIDENTIAL Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). In our 2023 Report, we observed that our prevalence testing showed that violent or disturbingcontent accounted for an extremely low percentage of PVP in August 2023. . A very lowpercentage of all Lenses submitted were found to include violent or dangerous behavior, whichwere all rejected. For this Report, it is worth noting that in Q3 2023, our ongoing monitoring identified an uptick inthe prevalence of violating content views for the violent and disturbing category. . In response,we launched new proactive detection mechanisms to target violent and disturbing content.Immediately after launch, the team were able to enforce significantly more proactively detectedSnaps daily. Consequently, we have reduced the prevalence of this category and we expect thisto decrease further As a result, we are pleased to report that prevalence for violent or disturbing content (as well asfor Dangerous Activities) has seen a further substantial fall in PVP to an extremely low category .This is a result of our specific efforts to reduce exposure to illegal and other violating contentfalling within our ‘Violent and disturbing content’ category. Our proactive content moderation hassuccessfully evolved to reduce the prevalence of violent and disturbing content. On Snapchat, our Transparency Report shows that Threats \& Violence still account for a relatively modest amount of all enforcement actions we took across all categories: 2.1% of all enforcementactions in the second half of 2023, compared with 2.6% in the second half of 2022. We receiveda slightly greater number of reports related to Threats \& Violence: 836,125 reports in the secondhalf of 2023, compared with 753,467 in the second half of 2022. We saw a slight decrease inconfirmed violations, with action against 114,315 of content and 83,743 accounts in the secondhalf of 2023 (compared with action against 167,811 pieces of content and 132,915 accounts in thesecond half of 2022). As a result of the low prevalence from our testing, and relatively consistent levels of reportingand enforcement, we continue to place the dissemination of content encouraging or engaging inviolent or dangerous behavior as falling within our extremely low likelihood category relative toother risks on Snapchat’s in-scope services. We continue to note that all of the risks we track onSnapchat have a relatively low prevalence compared to the prevalence of these issueselsewhere online and offline. 83 CONFIDENTIAL Severity Snap considers that the spectrum of “encouraging or engaging in violent or dangerous behavior”can vary considerably and covers a broad range of content types: ● Content relating to imminent, credible threats such as school or other mass shooting andbombing threats, although this is mainly a US-related risk and less relevant for EU users.Snap considers credible imminent threats to human life to constitute a severe harm. ● Viral “challenges” may cause injury (for example, the “Milk Crate Challenge” of 2021).Since well before the existence of social media, some people have sought out videos ofother people getting hurt. This content ranges from horrifying shock content, to relativelytame comedic pratfalls and minor injuries. Snap has assessed information published by governments and other third party sources andconsiders that if information encouraging or engaging in violent or dangerous behavior were tomaterialise on an online platform, these issues can vary considerably in severity, from our ‘severeharm’ category to our ‘significant harm’ category Overall, Snap considers content encouraging orengaging in violent or dangerous behavior to fall within our significant harm category. Overall potential risk prioritization Content encouraging or engaging in violent or dangerous behavior is one of the lowestlikelihood risk categories on Snapchat and runs the gamut from urgent, credible threats tohuman life which we continue to consider falls within our Level 1 overall potential risk (indeviation from our standard risk matrix), to unfortunate or even silly “fails” which we continue toconsider falls within our Level 3 potential overall risk prioritization. There is no change in thisassessment since our 2023 Report. Snap's Mitigations Highlights We devote significant resources to enforcing against truly harmful or shocking contentencouraging or engaging in violent or dangerous behavior. Snap’s Terms (in particular our Community Guidelines) address the dissemination of contentencouraging or engaging in violent or dangerous behavior. In addition, on public surfaces where we algorithmically recommend content, we take proactivemeasures to stop the dissemination of content encouraging or engaging in violent or dangerousbehavior. We use a mix of automation and human review to enforce our guidelines. As well as thecontent that is prohibited from Snapchat by our Terms above, content that relates to dangerous 84 CONFIDENTIAL challenges that is violating or harmful is also ineligible for promotion or recommendation on theDiscover and Spotlight surfaces of Snapchat. We have explained in Section 5.6 of the Report (Algorithmic Systems) the specific measures wehave taken to address risks including prevention of dissemination of content encouraging orengaging in violent or dangerous behaviour (such as dangerous challenges). In particular, wewould note that sensitive content distribution is limited on both Spotlight and Discover: ● In Spotlight, we limit the distribution of sensitive content based on the following rules: ○ We do not recommend sensitive content to users under 18 by default ○ We do not recommend sensitive content to new users . ○ For all other users, by default, we ensure the initial video watched in a session isnot sensitive and after that we ensure that sensitive content is only shownsparingly. ● In Discover, as in Spotlight, we limit the display of sensitive content for all users. We alsodo not show sensitive content to users under 18 by default and display of sensitivecontent can be disabled entirely in the Family Center. Any content anywhere on Snapchat can be reported in-app or on our web site, and “threats and violence” is one of the reporting reasons offered. Our Transparency Report shows that the median turnaround time for a “threats and violence” report is 27 minutes. We take additional measures to protect the well-being of the Snapchat community. For example,where our Trust and Safety and Law Enforcement Operations teams identify credible threats tohuman life, we have protocols in place for alerting local officials. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive moderation of content encouraging or engaging in violent or dangerous behavior and easy reporting. 85 CONFIDENTIAL Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our terms prohibit content encouraging or engaging in violent or dangerous behavior and we have provided a specific Threats, Violence \& Harm explainer which includes guidance on how we enforce this content. Our Transparency Report shows that the median turnaround time for a “threats and violence” report is 27 minutes. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove content encouraging or engaging in violent or dangerous behavior. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend content encouraging or engaging in violent or dangerous behavior i.e. there is no ‘violent or dangerous’ interest category. We mark certain shocking content as sensitive and limit the volume of recommendations. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, for example we have specific prevalence testing and monitoring moderation and enforcement data which we use to help detect and manage content encouraging or engaging in violent or dangerous behavior. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers who focus on child and digital safety. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the We are not working with other providers on violent and dangerous behavior specifically. However, we work with several groups in relation to child protection in general, including in relation to the new EU AAD Code. 86 CONFIDENTIAL crisis protocols referred to in Articles 45 and 48 respectively. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on harms (including a specific Threats, Violence \& Harm explainer) and how to get help in our Safety Center. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.29 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. General content authenticity measures Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion Dissemination of content encouraging or engaging in violent or dangerous behavior onSnapchat’s in-scope services is one of our lowest likelihood category risks. We recognize that thepotential harm arising from such content can be significant and we have therefore tracked thisrisk with an overall Level 3 potential risk rating. We devote significant resources to enforcingagainst truly harmful or shocking content encouraging or engaging in violent or dangerousbehavior as summarized above. Our prevalence testing shows the prevalence of this type ofcontent to be very low and failing on Snapchat’s in-scope services. As a result, we have concluded that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for dissemination of content encouraging orengaging in violent or dangerous behavior. Snap monitors this category to confirm whetherfurther mitigating measures might be required. There is no change in this conclusion since our2023 Report. 29 https://parents.snapchat.com. 87 CONFIDENTIAL 4.1.10 Dissemination of harmful false information Without mitigations, “Fake news,” (online) “disinformation” and “deep fakes” could conceivablybe present in videos published on Spotlight and For You, promoted in Stories on Public Profiles,in places and Snaps featured on Snap Map and in Lenses published via Lens Studio. Harmfulfalse advertising might include ads for content that mimics the appearance or function ofSnapchat features or formats or political advertising with false statements and slogans regardingimportant societal issues. Likelihood In practice, the dissemination of harmful misinformation is still not common on Snapchat. In our2023 Report we observed that our prevalence testing showed an extremely low prevalence of‘harmful false information’. We observed a significant further reduction in prevalence. As explained in our Transparency Reports, False information continues to account for only 0.1% of the total of all content enforced on Snapchat. This figure remained steady throughout bothhalves of 2022, as well as through the second half of 2023. We track Impersonation separately,and it similarly accounts for a very low percentage of our enforcement actions (0.3% in thesecond half of 2023, compared with 0.2% in the second half of 2022). Lenses with this type ofinformation are rarely submitted (only 0.0021% of all Lenses included misinformation in our 2023Report, all of which were rejected). As a result, we continue to place dissemination of harmful misinformation into our extremely lowlikelihood category relative to other risks.30 Severity Snap has assessed information published by governments and other third party sources andconsiders that if information encouraging or engaging in violent or dangerous behavior were tomaterialise on an online platform, it would fall within our significant harm category relative toother risks. Overall potential risk prioritization Harmful false information is classified as a Level 3 overall potential risk prioritization given itsextremely low prevalence on Snapchat, but is a risk that we apply significant resources tomitigate against, from the design of our platform to the ways we carefully review content before ithas an opportunity to reach a wide audience. There is no change in this assessment from our2023 Report. 30 This classification is also supported by the fact that Snapchat was not included in the report issued bythe European Commission: European Commission, Directorate-General for Communications Networks,Content and Technology, Digital Services Act – Application of the risk management framework to Russiandisinformation campaigns, Publications Office of the European Union, 2023, url. 88 CONFIDENTIAL Snap's Mitigations Highlights Snap has observed the negative societal effects of false information circulating on otherplatforms and is keen to avoid this being an issue on Snapchat. We devote significantenforcement resources to preventing the dissemination of harmful false information. Snap’s Terms address the dissemination of harmful false information. Under these policies, harmful false information includes false or misleading content that causes harm or is malicious. Itincludes impersonation, as well as disinformation, misinformation, malinformation, andmanipulated media that causes harm or is malicious, such as denying the existence of tragicevents, unsubstantiated medical claims, or undermining the integrity of civic processes. The design of the Snapchat app has been made to be hostile towards the dissemination of falseinformation. Our platform design architecture makes it difficult to spread misinformation.Snapchat has made conscious design decisions to restrict the ability for content to go viral,including limiting the remix functionality and applying short default retention to content. Onsurfaces where a broader audience can potentially be reached our proactive detection makes itdifficult for misinformation to reach a large audience. Moreover, our content platform, Discover,features content from approved media publishers and content creators. Our entertainmentplatform for user-generated content, Spotlight, is proactively and a priori moderated beforecontent can reach a wide audience. On Snapchat where we algorithmically recommend content, we also take proactive measures tostop the dissemination of false information. When we consider whether to allow content for algorithmic recommendations, we apply additional rules. Content is “Not Eligible for Recommendation” when it contains misleading or sensationalized headlines. Also of note: thereare distinctions between the content that may appear on the 4th tab (“Stories”) and 5th tab(“Spotlight). In Discover, we allow political content, but only from trusted media partners, SnapStars, and certain popular accounts, to be algorithmically recommended beyond theirsubscribers. In Spotlight, we do not allow political content from anyone; this surface ispre-moderated to prevent political content from achieving reach. Beyond content from media partners and users, misinformation may come in the form ofadvertising. Every political, health or sensitive issue ad is reviewed by humans on the ad reviewteam. Our ad policies require that these advertisers provide supporting documentation for allclaims. We reject ads that contain unsubstantiated or false claims. Our ad policies also requirethat advertisers be transparent about the paying entity; this information is displayed to the enduser in the “slug” onscreen to prevent advertisers from impersonating other entities. All political ads are logged in our political ads library. 89 CONFIDENTIAL As explained in our Moderation and Enforcement sections, we use a mix of automation (such asabusive language detection, image recognition models, and account history) and human reviewto prevent and remove content violating our Terms relating to harmful false information. Ourhuman review teams maintain training examples of recurring harmful false information; they arefamiliar with the most common unfounded conspiracy theories that circulate online. When theyencounter new false or ambiguous information that may relate to politics, health, or tragic events,they fact check using trusted resources. Where necessary, they escalate emergent narratives tothe Platform Policy team for review. Where misinformation is being spread from an account thathas been taken over or is falsely claiming to represent someone, our team works tirelessly torestore accounts to their rightful owners, and to remove accounts or content that deceives othersabout one’s identity. Any content anywhere on Snapchat can be reported in-app or on our web site, and “false information” and “Impersonation” are two of the reporting reasons offered. Our Transparency Report shows that the median turnaround time for a “false information” report is 12 minutes and for an “impersonation” report is 3 minutes. When considering impersonation, we allow parodythat is unlikely to cause confusion; when reviewing content, our teams are trained to distinguishbetween these permissible activities and harmful impersonation attempts.31 During election seasons, such as the recent European elections, we contract third-partyfact-checking organizations, such as Poynter, to support our work. We have also partnered withnumerous governments around the world to inform Snapchatters about upcoming elections andto vote. Regarding the European elections specifically, we consider that these unfolded in apositive online environment with no major threats overall. This was confirmed by the EuropeanCommission and independent observers, who confirmed that they did not observe major onlinethreats. Snap saw a small uptick in reported activity, but did not receive or observe any materialincidents or threats. Our moderation and reporting tools worked well and none of the reportedpieces of content were verified as misinformation on Snapchat. In the lead up to the elections,Snap attended multiple cross functional stakeholder meetings, including with civil societyorganizations, regulators, including the European Commission, and other platforms to shareinformation. We believe these stakeholder meetings contributed to the positive outcome, and welook forward to continuing these engagements. Snap sent a push notification to over 50 millionusers to urge them to vote in the elections, and made AR election lenses available to promotecivic engagement. We are proud to have done our part to contribute to the highest observed 31 Correctly and consistently enforcing against false information is a dynamic process that requiresup-to-date context and diligence. As we strive to continually improve the precision of our agents’enforcement in this category, we have chosen, since H1 2022, to report figures in the "Content Enforced"and "Unique Accounts Enforced" categories that are estimated based on a rigorous quality-assurancereview of a statistically significant portion of false information enforcements. Specifically, we sample astatistically significant portion of false information enforcements across each country and quality-check theenforcement decisions. We then use those quality-checked enforcements to derive enforcement rates witha 95% confidence interval (+/- 5% margin of error), which we use to calculate the false informationenforcements reported in the Transparency Report. 90 CONFIDENTIAL turnout of the last 30 years, with 51.08% of the 357 million eligible citizens participating in theelection. We published a blog post detailing our measures and experiences.32 Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat is not an attractive platform for spreading misinformation, in particular because it is difficult to reach a broad audience and content deletes by default. Snap has made conscious design decisions to restrict the ability for content to go viral and limiting the remix functionality to specific content types and applying short retention to content. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our terms prohibit misinformation. We have a specific Harmful False or Deceptive Information explainer which explains our approach to enforcement. Our Transparency Report shows that the median turnaround time for a “false information” report is 12 minutes. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, we use specific proactive and reactive moderation procedures to prevent and remove misinformation. In particular, Discover features content only from approved media publishers and significant content creators. Our entertainment platform for user-generated content, Spotlight, is proactively and a priori moderated before content can reach a wide audience. 32 https://newsroom.snap.com/snap-eu-election. 91 CONFIDENTIAL Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend content encouraging or engaging in misinformation i.e. there is no ‘misinformation’ interest category. We take steps to prevent content with misleading or sensationalist headlines. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Every political, health or sensitive issue ad is reviewed by humans on the ad review team. We reject ads that contain unsubstantiated or false claims. All political ads are logged in our political ads library. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, for example we have specific prevalence testing and monitoring moderation and enforcement data which we use to help detect and manage harmful false misinformation. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers, our trusted flaggers may also report misinformation, but this rarely happens because of the limited amount of misinformation on the platform. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Snap has not yet signed up to be a member of the EU disinformation code. We have limited exposure to the risk and use our limited resources to focus on other codes relating to risks more relevant to Snapchat’s in-scope services. However, Snap works closely with French regulator Arcom, which monitors industry action against misinformation. We have also worked closely with the Commission and other stakeholders during the recent EU elections. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. This includes a specific Harmful False or Deceptive Information explainer. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting; and we provide guidance to parents on the 92 CONFIDENTIAL control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. web. Our new parents site provides additional guidance for parents and carers on risks and support.33 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. We recognise the risk that generative AI could be used to generate harmful false misinformation, including deep fakes. Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion We recognise a risk of significant harm that could arise from harmful false information. In practice,Snapchat’s in-scope services have very little exposure to harmful false information. It is one of ourlowest likelihood categories of risks. As a result we track this risk as an overall significantpotential risk. Snapchat has significant measures in place to prevent harmful misinformation, inparticular the design and function of Snapchat’s in-scope services which limits the spread ofcontent, limits the places where user generated can reach a broader audience and targetsproactive moderation at those areas to prevent harmful misinformation from becoming viral. We have a rapid response time when harmful false information does slip through. Our Transparency Report shows that the median turnaround time for a “false information” report is 12 minutes and for “impersonation” is 3 minutes. We have also observed further significant falls in the prevalenceof harmful false information from our prevalence testing. As a result, we have concluded that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for dissemination of harmful falseinformation. There is no change in this conclusion from our 2023 Report. 4.1.11 Dissemination of fraud and spam Without mitigations, content encouraging or engaging in violent or dangerous behavior couldconceivably appear on Snapchat’s in-scope services. 33 https://parents.snapchat.com. 93 CONFIDENTIAL Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). In our 2023 Report: ● We observed from our prevalence testing that Fraud and Spam had a lowprevalenceAugust 2023. We are happy to confirm in this Report that Fraud and Spam continue to beon a downward trajectory confirming that ongoing detection and enforcementenhancements have been having the desired impact resulting in a reduction inprevalence. There was a substantial further decrease in prevalence in July 2024. ● Our Transparency Report shows that “Spam and Fraud” reports continue to lead to a moderate volume of content and account enforcements on our content online platforms,such as Spotlight and Discover. In 2023 we started to see an influx of Spam accounts.Consequently, we significantly increased Fraud and Spam enforcement, which jumpedfrom 657,077 in the second half of 2022 (8.2% of all enforcement actions) to 1,380,341 inthe first half of 2023 (22.2% of all enforcement actions). This started to decrease in thesecond half 2023 where the number of enforcements dropped to 1,002,278 (18.6% of allour enforcement actions). The project was able to successfully complete in December2023. In addition, evidence34 submitted to a UK Home Affairs Committee inquiry35 into fraud by AlisonThewliss MP (SNP, Glasgow Central) quoted figures from TSB on the prevalence of fraud ondifferent platforms. She stated that 70% of the frauds that TSB was picking up were beingperpetrated on Meta—24% on Facebook and 46% on Instagram—4% on Snapchat and 23%across other platforms. This is further evidence that indicates that Fraud has low prevalence onSnapchat in general. Overall, given the further reductions in prevalence of Spam and Fraud content on Spotlight andDiscover, we consider the dissemination of fraud and spam in terms of relative likelihood onSnapchat’s content platforms to fall within the Extremely Low likelihood category. This is achange from our conclusion in our 2023 report where we considered that Spam and Fraud fellwithin the very low likelihood category. We observed that fraud was by far the most common reason for advertising on Snapchat to berejected between 1 July 2022 and 30 June 2023. This trend has continued this year, As a result, 35 UK House of Commons Home Affairs Committee, Committee Press Release. 34 UK House of Commons Home Affairs Committee, Home Affairs Committee. Oral evidence: Fraud, HC125, Wednesday 22 November 2023, url. 94 CONFIDENTIAL we consider dissemination of fraud and spam to be the low likelihood risk for our advertisingsystems, relative to other risks. Severity Snap has assessed information published by governments and other third party sources andconsiders that, if spam and fraud were to materialise on an online platform, it would fall within our “significant harm” category relative to the other risks we track. Overall potential risk prioritization We consider Fraud and Spam to fall within the Level 3 overall potential risks category comparedto other more severe harms. We devote significant resources to protecting our users from fraudand spam in user-generated content and advertising. There is no change in this assessmentsince our 2023 Report, save to reflect the fact that Fraud and Spam in relation to content is nowin the Lowest Relative Likelihood category. Snap's Mitigations Highlights Snap devotes enforcement resources to preventing the dissemination of content that includesfraud and spam. Snap’s Terms address the dissemination of fraud and spam across the full range of servicerecipients that create content on Snapchat: users, media partners and advertisers. When we consider whether to allow content for algorithmic recommendations, we apply additional rules. Commercial Promotions For commercial promotion within content from media partners or users, we apply our Commercial Content Policy. The Commercial Content policy also prohibits Deceptive Content. The Commercial Content Policy also outlines rules to protect Snapchatters from potentially misleadingreferences to Snap. Commercial content must not suggest an affiliation with or endorsement bySnap or its products. Advertising Snap’s Advertising Policies detail the criteria that our automation and human review teams apply while considering whether to allow or reject an ad on our platform. Our advertising policies prohibit Deceptive Content. The advertising policies for financial products and services add further detail about the kind of deceptive content that is prohibited. 95 CONFIDENTIAL As explained in Section 5 of this Report, we are vigilant in our moderation and enforcement ofour Terms, including against fraud and spam content and ads. On Snapchat’s in-scope serviceswhere we algorithmically recommend content, we take proactive measures to stop thedissemination of fraud and spam. We use a mix of automation (such as abusive languagedetection, image recognition models, and account history) and human review to moderate andenforce our guidelines. Content on Snapchat can be reported in-app or on our Support Site, and “spam and fraud” is one of the reporting reasons offered. Our Transparency Report shows that the median turnaround time for a fraud and spam content report is 2 minutes. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat is not an attractive platform for spreading fraud and spam, in particular because it is difficult to reach a broad audience, and Snapchat has made conscious design decisions to restrict the ability for content to go viral. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our terms prohibit fraud and spam, with specific rules for commercial content and advertising, and we strictly enforce these rules. Our Transparency Report shows that the median turnaround time for a “fraud and spam” report is 2 minutes. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant Yes, we use specific proactive and reactive moderation procedures to prevent and remove misinformation. In particular, Discover features content only from approved media publishers and significant content creators. Our entertainment platform for user-generated content, Spotlight, is proactively and a priori moderated before content can reach a wide audience. 96 CONFIDENTIAL decision-making processes and dedicated resources for content moderation. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend fraud or spam, i.e. there is no ‘fraud’ or ‘spam’ interest category. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Advertising is subject to moderation before publication, with most ads subject to a human review. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we have specific transparency reports for fraud and spam. Prevalence testing is generally not used for ads since they are prescreened and there is a higher bar for bad actors for ads since it requires payment configurations. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers, our trusted flaggers may also report fraud spam, but this is not generally the focus of their efforts. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. We are not currently members of a dedicated group or code addressing the issue of fraud and spam online. However, we are members of several organizations and trade associations that tackle online issues facing the industry. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.36 Content Authenticity Generalcontent authenticity measures. Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or 36 https://parents.snapchat.com. 97 CONFIDENTIAL Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion We address Fraud and Spam as a very low likelihood issue in general, and low likelihood for ouradvertising systems, relative to other risks faced by Snapchat’s in-scope services. Given the riskof significant harm arising from fraud and spam, we categorize this issue overall as a Level3potential risk prioritization. We handle significant volumes of enforcement and rejections everymonth. Our prevalent testing shows this is working, with significant further reductions in theprevalence of fraud and spam on public content surfaces of Snapchat. Ad rejection rates haveremained consistent with the overall increase in ads reviewed. As a result, we have concluded that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for fraud and spam. We are pleased with thefurther reductions in the prevalence of fraud and spam on Snapchat’s public content surfaces.Snap will continue to monitor this category to ensure the consistently higher proportion of adrejections for fraud and spam are not an indication that further mitigating measures might berequired with regards to ads. This amounts to a slight change of focus from the conclusion inour 2023 Report. 4.1.12 Dissemination of information related to other illegal activities As we allow users to publish content, we recognise that without mitigations it is possible thatinformation related to other illegal activities not already captured by our other categories abovemay be disseminated on Snapchat’s in-scope services. Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). Our prevalence measurement and transparency reporting continue to track the prevalence of known significant issues that could potentially impact online platforms, including Snapchat, as 98 CONFIDENTIAL informed by our work with Trusted Flaggers, industry groups and our safety advisory board andinternal cross functional working groups. These categories are already addressed above, as inour 2023 Report, and we are not currently aware of other significant issues. With the introduction of the Digital Services Act, we introduced a new reporting option to report‘illegal content’ in general. We have seen very few reports being made through this reportingoption and, when it is used, the quality of the reports are very low (for example, missing keyinformation to be able to identify the content in question and/or the illegal nature of the activity orthe report concerned harmless activity) and are usually not actionable. With regards to the verysmall number of reports that were actionable, almost all of these related to our existing illegal andother violating content categories referred to above. We have not observed any significant newillegal activity categories. As a result we still believe the dissemination of information related to other illegal activities to fallwithin the Extremely Low likelihood category relative to other risks identified by Snap. Severity The extent of harm that might be risked by information relating to other illegal activity woulddepend on the issue. Snap has specific categories for risks concerning the dissemination ofinformation which are most relevant to online platforms. As a result, we categorize the risk ofharm in general from information relating to other illegal activity within our ‘significant harm’category. Snap would consider the issue of illegal activity to fall within our ‘severe harm’category where the content includes a credible threat to human life, safety, or well-being. Overall potential risk prioritization In general Snap assesses the overall potential risk of the dissemination of this type of content tobe Level 3 i.e. Snap’s lowest risk category compared to other risks. As in other cases, where anyissue arises that poses an imminent and credible threat to human life, safety, or well-being, Snaptreats this issue with a Level 1 overall potential risk. There is no change in this assessment sinceour 2023 Report. Snap's Mitigations Highlights Snap is sensitive to the issue of internet platforms being used to engage in illegal activity ingeneral. We believe the steps Snap has taken to mitigate harm in general have substantiallydiminished the likelihood that Snapchatters will find other illegal activity on our platform beyondthe categories assessed in the previous dissemination issues discussed above. Unlike many ofour peers, Snap does not offer an open news feed where unvetted publishers or individuals have 99 CONFIDENTIAL an opportunity to broadcast illegal content, nor does Snapchat offer a ‘reshare’ functionality thatwould encourage virality, and does not allow user-generated content to gain wide viewershipwithout going through human review. Snap’s Terms prohibit users from posting content that’s illegal in their jurisdiction or using Snapfor any illegal activity. Snap’s Community Guidelines and enforcement strategy are driven bySnap’s values and desire to facilitate a fun, positive user experience. Snap’s CommunityGuidelines therefore prohibit both illegal activity and activity which Snap considers harmful oragainst our values, but which is not necessarily illegal under EU law. For this reason, Snap is likelyover-inclusive on its policies against illegal activity. It is possible, despite Snap’s terms and policies prohibiting such practices, as well as Snap’smoderation and enforcement mechanisms, that malicious actors will find ways to circumventSnap’s enforcement mechanisms and practices in order to engage in illegal activity, which couldthen appear on Snap’s public surfaces. Snap removes illegal content and activity as we becomeaware of it, cooperates with law enforcement, and disables the accounts of egregious or repeatviolators. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive moderation for some other illegal activities. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our terms prohibit other illegal activities and they are strictly enforced. Our legal team, supported by external counsel as needed, reviews reports of new issues to confirm illegality and appropriate enforcement action. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, Yes, general proactive and reactive moderation procedures to prevent and remove other illegal activities. 100 CONFIDENTIAL where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend information relating to illegal activity i.e. there is no ‘illegal activity’ interest category. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. We rely on our Trusted Flaggers, industry groups and our safety advisory board and internal cross functional working groups to ensure we are prioritizing the right issues. With the introduction of the Digital Services Act, we have introduced a new reporting option to report ‘illegal content’ in general, and we expect to use data gathered from this option to provide us with greater visibility on the prevalence of information relating to other illegal activity on Snapchat. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers who are able to flag other illegal activities. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups on prominent issues facing online platforms. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. Protection of Minors Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting; and we provide 101 CONFIDENTIAL Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.37 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. General content authenticity measures: Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion We prohibit the dissemination of information relating to illegal activities and criminal activity in our Terms. We specifically track the issues relating to the dissemination of information which we consider to have the most relevance to online platforms, such as Snapchat. We treat otherdissemination issues as a Level 3 overall potential risk compared to other harms and have takensteps to mitigate these risks. We regularly review our risk categories using our Risk Detection andManagement processes. We have seen few reports using our new option to report ‘other illegalactivity’ and have seen any new categories emerge as the vast majority are either not actionableor relate to one of our existing categories. As a result, we have concluded that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for information relating to other illegalactivities. Since our 2023 Report, we have monitored DSA enquiries, including to our newreporting option, but have so far not identified any new trends or the need to establish newillegal or harmful content categories. 4.2 Category 2: Negative Effects on Fundamental EU Rights (Article 34.1.b / DSA Recital 81) In this part of the Report, we explain the results of our assessment on actual or foreseeablenegative effects of Snapchat’s in-scope services on our Fundamental EU Rights as required by 37 https://parents.snapchat.com. 102 CONFIDENTIAL Article 34.1.b and Recital 81 of the Digital Services Act. Those Fundamental EU Rights are set outin the Charter of Fundamental Rights of the European Union (the “Charter”)38. We have assessedin particular the rights to human dignity, freedom of expression and of information, includingmedia freedom and pluralism, private life, data protection, non-discrimination and consumerprotection. We also consider the rights of the child, including how easy it is for Teens tounderstand the design and functioning of the service, as well as how Teens can be exposedthrough their service to content that may impair Teens’ health, physical, mental and moraldevelopment. Such risks may arise, for example, in relation to the design of online interfaceswhich intentionally or unintentionally exploit the weaknesses and inexperience of Teens or whichmay cause addictive behavior. Category 2 - Negative effects on Fundamental EU Rights Category Relative likelihood of risk occurring on Snapchat Harm classification industry wide Risk Prioritization Snap’s mitigations 4.2.1 Right to human dignity Extremely Low Likelihood Severe harm industry wide Level 1 Low Risk / Reasonable, proportionate and effective 4.2.2 Right to freedom of expression Extremely Low Likelihood Significant harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective 4.2.3 Right to private life Extremely Low Likelihood Serious harm industry wide Level 2 Low Risk / Reasonable, proportionate and effective 4.2.4 Right to data protection Low Likelihood Severe harm industry wide Level 1 Low Risk / Reasonable, proportionate and effective 4.2.5 Right to non-discrimination and freedom of religion Extremely Low Likelihood Serious harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective 4.2.6 Children’s rights Extremely Low Likelihood Severe harm industry wide Level 1 Low Risk / Reasonable, proportionate and effective and we are actively participating in efforts to develop an EU wide guidance to assess if further industry 38 Charter of Fundamental Rights of the European Union, url. 103 CONFIDENTIAL measures are needed. 4.2.7 Right to consumer protection Extremely Low Likelihood L Significant harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective 4.2.8 Right to property N/A. Already covered under Category 1: Dissemination of content that infringes on intellectual property rights 4.2.1 Right to human dignity All public spaces displaying user generated content have the potential for the dissemination ofcontent that may undermine human dignity. We recognise that without mitigation such contentcould conceivably appear in any of Snapchat’s in-scope services displaying user generatedcontent, from videos featured on Spotlight / Discover, to Place Stories on Snap Map. Advertisingcould, for example, include hate speech or discriminatory elements. All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). Snapchat, as with other platforms that host user generated content, may be used to spreadcontent that undermines respect for human dignity. Without mitigations, this could includecontent that promotes: ● Human trafficking and/or the sale of coerced sex; ● Child sexual abuse material; ● Terrorism; ● Self-harm, including the promotion of self-injury, suicide or eating disorders; ● Incitement to violence or hatred directed against a group of persons or a member of agroup based on any of the grounds referred to in Article 21 of the Charter. As Snap takes these issues very seriously and has implemented several levers to prevent thiscontent from being distributed on the platform Likelihood We have assessed the relative likelihood of Snapchat’s inscope services disseminating contentthat may undermine human dignity, based on Policy Violating Prevalence (PVP) via randomsampling and our transparency report data in Section 4.1 of this Report, as follows: Category Relative likelihood of risk occurring on Snapchat 104 CONFIDENTIAL 4.1.1 Dissemination of child sexual abuse material Extremely Low Likelihood 4.1.2 Dissemination of illegal hate speech Extremely Low Likelihood 4.1.4 Dissemination of terrorist content Extremely Low Likelihood 4.1.6 Dissemination of adult sexual crimes Extremely Low Likelihood 4.1.8 Dissemination of content that glorifies self-harm, including the promotion of self-injury, suicide or eating disorders Extremely Low Likelihood We therefore continue to assess that the relative likelihood that the in-scope services ofSnapchat would have an actual or foreseeable negative effects of the right to human dignity fallswithin our Extremely low likelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that if content that may undermine human dignity were to materialise on an onlineplatform, it would fall within our ‘severe harm’ category. We have assessed the severity of harm caused by each of these categories of content thatsignificantly undermines the right to human dignity in Section 4.1 of this Report, as follows: Category Relative likelihood of risk occurring on Snapchat 4.1.1 Dissemination of child sexual abuse material Severe harm industry wide 4.1.2 Dissemination of illegal hate speech Significant harm industry wide 4.1.4 Dissemination of terrorist content Serious harm industry wide 4.1.6 Dissemination of adult sexual crimes Serious harm industry wide 4.1.8 Dissemination of content that glorifies self-harm, including the promotion of self-injury, suicide or eating disorders Serious harm industry wide Our assessment shows a variety of harm ranging from significant to the most severe. We continueto choose to assess the category of actual or foreseeable negative effects of the right to humandignity using the highest severity rating of the categories we have assessed. Given the outsizepotential for harm for some of the risks to human dignity, Snap continues to consider this risk tofall within the severe risk of harm category. 105 CONFIDENTIAL Overall potential risk prioritization Although the prevalence of content that negatively impacts users’ rights to human dignity hasbeen assessed to be in the lowest likelihood category, we have assessed the severity of this riskto be severe. As a result, overall, we consider the negative effects on the right to human dignityto fall within the Level 1 overall potential risk prioritization category. As described in our riskmethodology in Section 1, we assess overall potential risk on a case by case basis and Snapreserves the option to deviate from the overall potential risk matrix we use as a guide. This is oneof the cases where we have chosen to deviate. There is no change in this assessment from our2023 Report. Snap's Mitigations Highlights Snap’s approach to protecting users’ rights to human dignity and mitigating the related risks isimplemented through a robust framework of content moderation as described in the Moderationsection. Activity that may undermine human dignity is not permitted on our platform under our Terms and Community Guidelines. We have tools within the app where individuals may report this type of activity to our Trust \& Safety team. They will then investigate the report and take action. Inthe event the report concerns any imminent threat to life, Snap will alert the appropriateauthorities. We also maintain relationships with several entities on a global basis through ourTrusted Flagger program and they may also report activity to our Trust \& Safety team. TheseTrusted Flaggers are vetted and they possess an expedited means for contacting our teams. In addition to effective content moderation, Snap has additional mechanisms in place to enhancethe right to human dignity for users. For example, because Snapchat is a platform designed forcommunications between real friends, it can play a unique role in empowering friends to helpeach other through difficult times. When our Trust \& Safety team recognizes a Snapchatter indistress, they can forward self-harm prevention and support resources, and notify emergencyresponse personnel when appropriate. The resources that we share are available on our globallist of safety resources, and these are publicly available to all Snapchatters. Specific Mitigations This table lists a number of specific mitigations to address risks to human dignity on Snapchat’sin-scope services. To avoid duplication, this table includes cross-references to other sections ofthis Report. In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in this 106 CONFIDENTIAL Report, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive moderation for CSEAI and other illegal content that undermines human dignity. We also have tools within the app where individuals can report this type of activity to our Trust and Safety team. When our Trust and Safety team recognizes a Snapchatter in distress, they can forward self-harm prevention and support resources, and notify emergency personnel when appropriate. For example, if a user searches for suicide related terms we will surface our Here For You tool. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our Terms prohibit CSEAI and other illegal content that undermines human dignity and they are strictly enforced. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent CSEAI and other illegal content that undermines human dignity. We have terms in place to prevent Media Partners from publishing illegal or harmful content on Discover. All Media Partners are vetted prior to being permitted to distribute their content broadly on Snapchat by a team of editors. Media Partners go through an editorial review of their content, a reputational search (to evaluate if a Media Partner has a history damaging press, legal actions, etc.), and compliance review before they’re able to distribute content. Media partners are proactively moderated, and the content of their shows/editions are reactively moderated. Senior partner managers will relay feedback to Media Partners to remove or change content. If a partner refuses, we could just remove it ourselves, but partners typically comply. 107 CONFIDENTIAL Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not knowingly recommend content that would negatively affect the right to human dignity i.e. there are no interest categories that we consider to negatively affect human dignity. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we have specific prevalence testing and transparency reports for CSEAI, terrorist content, and other illegal content that undermines human dignity. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to CSEAI and other illegal content that undermines human dignity. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups e.g. EUIF. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to get help in our Safety Center. For example, if a user searches for suicide related terms we will surface our Here For You tool. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.39 Content Authenticity Some content authenticity measures have been taken in respect of content that 39 https://parents.snapchat.com. 108 CONFIDENTIAL Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. undermines human dignity. Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion Snap considers risks to human dignity to have a Level 1 overall potential risk. In response it hasput in place a range of mitigation measures. This includes in particular our proactive contentmoderation which is designed to detect and prevent CSEAI from appearing on each ofSnapchat’s in-scope services. For example, our automated and human review on Spotlight. Ourprevalence testing has allowed us to improve this proactive content moderation. As a result,we’ve reduced the prevalence of CSEAI and other content that may undermine human dignity on Snapchat’s in-scope services to the lowest likelihood level. See Section 4.1. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for risks to human dignity. There is nochange in this conclusion from our 2023 Report. 4.2.2 Right to freedom of expression and assembly Snapchat is an app whose mission is to empower people to express themselves, live in themoment, learn about the world, and have fun together. By design, the app itself presents anopportunity to enhance the freedom of expression and assembly of Snapchatters. However,without mitigations, Snap, alongside other digital platforms hosting user-generated content,presents some risk to these rights and freedoms. These risks could include: Algorithmic biases,content moderation bias, submission of abusive notices and self-censorship. Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). It is difficult to quantify the likely risk of negative impact on freedom of expression and assembly.Algorithmic biases and self-censorship are difficult to detect. We rely on user feedback and 109 CONFIDENTIAL testing to flag significant incidents. At present, we are not aware of any significant bias ofself-censorship issues in the algorithms used by Snapchat’s in-scope services. We continue tomonitor the number and nature of the general community support requests we receive and thisdata does not identify any trend that suggests Snapchat may be negatively impacting freedom of expression or assembly. Our transparency reports continue to show that in general we receive low incidents of illegal content reports from recipients of Snapchat or authorities where we chosenot to take enforcement actionBased on the lack of reporting Snap has received and the overalldesign of Snapchat (which does not generally provide a platform for political public content ingeneral), we deem this to fall within the Extremely low likelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that if content that may undermine human dignity were to materialise on an onlineplatform it would fall within our significant harm category. However, Snapchat generally is not aplatform for political or activist content and so the impact on freedom of expression and assemblyis unlikely to be severe on Snapchat compared with other spaces on the internet dedicated tosuch content. Overall potential risk prioritization Although it can be difficult to determine, the lack of reporting and Snap’s overall design, indicatesthat the prevalence of issues relating to freedom of expression and assembly are low. As Snapchat’s in-scope services do not generally amplify political or activist public content, theseverity of any freedom of expression risk is significant but not serious or severe. We considerthat freedom of expression risks fall within the Level 3 category overall. There is no change inthis assessment from our 2023 Report. Snap's Mitigations Highlights Our Terms clearly define certain topics which we prohibit, including false information thatthreatens public health (e.g. COVID-19 vaccinations), civic processes, or that denies tragic events(like the Holocaust). We also have an explainer to help our community understand how we handleharmful false or deceptive information. This provides clarity on the limits we have when it comesto freedom of expression and assembly. Our platform is generally not a place for political or activist public content. Such content is noteligible for promotion on Spotlight and user content on Discover is only from a small number ofpopular, entertaining community creators and their content is moderated by humans against ourContent Guidelines. All Media Partners are vetted prior to being permitted to distribute their 110 CONFIDENTIAL content broadly on Snapchat by a team of editors. These Partners include news organizations,which are subject to their own professional rules. Media Partners go through an editorial reviewof their content, a reputational search (to evaluate if a publisher has a history damaging press,legal actions, etc.), and compliance review before they’re able to distribute content. As a result,we provide a balanced approach to political and activist public content on Snapchat that isdesigned to limit the sources of such information to professional media partners. Snapchat utilizes content moderation policies and systems to protect users’ rights to freedom ofexpression and access to accurate information. As all of our user generated content is moderatedby a mix of automation and human moderation, we proactively remove content that does notmeet our policies before being broadly distributed. In some cases, content against our policiesmay make it past moderation by mistake. In those cases, we rely on Snapchatters to report thecontent for re-moderation. As explained when discussing the dissemination of content thatinfringes on intellectual property rights, Snap respects the doctrine of “fair use,” i.e., that there arecertain circumstances (such as news reporting, social commentary on issues of public interest,criticism, parody, or education) where excerpts of copyrighted material could be distributedwithout permission from or payment to the copyright holder. This helps reinforce the rights offreedom of expression and the freedom of assembly. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive moderation to protect users’ rights to freedom of expression and access to accurate information. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, terms prohibit harmful false or deceptive information and they are strictly enforced. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where Yes, specific proactive and reactive moderation procedures to protect users’ rights to freedom of expression and access to accurate information. 111 CONFIDENTIAL appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. On Snap Map, our editorial oversight protection for content showing up on Snap Map tries to strike the right balance between the need to preserve the public safety versus the free flow of information and expression. Examples of this include in February of last year when Russia moved into Ukraine, Snap Map developed tooling that allowed us to block all of Ukraine from creating content. This was in response to concerns that Russia was using it for their own strategic purposes (propaganda and tracking the movement of Ukrainians). Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, the pool of content recommended by our algorithmic systems does not generally include political or other important societal matters. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we consult with our safety advisory board to ensure Snapchat is set up appropriately and monitor community reports for issues relating to freedom of expression. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. No, we do not work with trusted flaggers for users’ rights to freedom of expression and access to accurate information. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, our Crisis Protocols handle issues related to users’ rights to freedom of expression and access to accurate information. We have recently exercised these protocols successfully during the French riots in June 2023. Note, we will continue to reassess and explore the opportunity to join the EU disinformation code. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms (including harmful false or deceptive information), moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. 112 CONFIDENTIAL Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to allow Teens to express themself without the pressures of friends lists, comments and likes. We have community, ad and content guidelines that are specific to teens. We also offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.40 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Yes, Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. Conclusion Snap continues to consider the overall risk to be within the level 3 risk prioritization categorygiven the stakes and the severity of threats to freedom of expression, despite low prevalenceand robust protections in place. Snap’s mission is to be an expressive platform where users canbe their authentic self, and we view our obligation to facilitate freedom of expression asfoundational. While harms to freedom of expression are hard to detect, and we are not aware ofany significant bias of self-censorship issues in the algorithms used by Snapchat’s in-scopeservices, we provide avenues for our users to report these issues to us, and we value andrespect user feedback. We continually evaluate and evolve our algorithms, including to reduceperceived biases, and monitor for and respond to events that could impact freedom ofexpression. We couple this with enforcement of our Terms and our robust moderation practicesto provide a platform where users feel free to express themselves in the world. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures to address risks relating to freedom ofexpression. Snap monitors its impact on this fundamental right category to confirmprevalence continues to decline, or whether further mitigating measures might be required.There is no change in this conclusion from our 2023 Report. 40 https://parents.snapchat.com. 113 CONFIDENTIAL 4.2.3 Right to private life We understand well that online platforms can be used to spread content that undermines respectfor private and family life, and that such content can have traumatic consequences if not properlymitigated. On Snapchat, without mitigations, content that undermines private and family life andpersonal data privacy could conceivably appear in any of Snapchat’s in-scope services displayinguser generated content, including information in videos featured on Spotlight / Discover andSnap Map. Snapchat’s platform architecture, combined with its commitment to responsible policyenforcement across our content surfaces, establishes safeguards against negative impacts to theprivate life of users. Likelihood In our 2023 Report, we explained that: ● Our prevalence testing showed that “invasion of privacy” made up an extremely lowpercentage of Policy Violating Prevalence on Snap in August 2023 (see our PrevalenceTesting chapter). ● Snap also receives low numbers of privacy-related queries from recipients. As a result,we deem this risk to fall within our Extremely Low Likelihoodlowest relative likelihoodcategory. We have been pleased to observe that this situation has continued: ● Our prevalence testing showed that “invasion of privacy” had seen a further, substantialfall in prevalence. It is now at a very low level. We therefore continue to assess that the relative likelihood that the in-scope services ofSnapchat would have an actual or foreseeable negative effects of the right to private life fallswithin our Extremely low likelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that if content that may undermine private and family life and personal data privacywere to materialise on an online platform it could fall within our ‘serious harm’ category. Overall potential risk prioritization Given the stakes and the severity of threats to private life, Snapchat assesses the overall risk tobe within the Level 2 category, despite low prevalence and robust protections in place. Asdescribed in our risk methodology in Section 1, we assess overall potential risk on a case by casebasis and Snap reserves the option to deviate from the overall potential risk prioritization matrixwe use as a guide. This is one of the cases where we have chosen to deviate. There is nochange in this assessment from our 2023 Report. 114 CONFIDENTIAL Snap's Mitigations Highlights Snap takes a multifaceted approach to mitigating negative impacts to private life and personaldata protection, starting from the way Snap develops its own enforcement mechanisms. Privacyis the first of Snap’s four core platform governance values, which remains paramount as Snapcontemplates the development of supplementary enforcement mechanisms that could potentiallyimpact users' personal data. Through Snap’s Platform Governance Framework, efforts to mitigateor understand harm must advance one or more of the platform governance values, and beconsistent with the balancing principles of necessary, proportional, and legitimate. The principlesof necessity, proportionality, and legitimacy derive directly from established human rights law andjurisprudence, and have been adapted for application in many different contexts, including asguiding principles for safeguarding against digital surveillance. Incorporating these principles intoSnap’s framework anchors our approach in an internationally validated, rights-respectingmethodology––one that is familiar to, and utilized by, policymakers and advocates in everyregion of the world. In developing this framework, we’ve drawn on a large body of principles andexpertise from across the digital policy, human rights, and online privacy communities. We also mitigate these risks through intentional product design choices. Privacy by Design isSnap’s approach to building products that consider user privacy from inception. Each product issubject to a PASS Review (Privacy Assessment System) to ensure that our products do not misuseuser-data. We also engage with experts in the privacy and human rights community to inform ourapproach. This includes collaborations and engagement with individual experts (such as experton human rights, privacy, and online safety Brittan Heller and former ICO Commissioner ElizabethDenham, and several others), as well as think tanks and research collaborations. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Privacy by Design is Snap’s approach to building products that consider user privacy from inception. Each product is subject to a PASS Review (Privacy Assessment System) to 115 CONFIDENTIAL ensure that our products do not misuse user-data. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, for example, our Community Guidelines prohibit impersonation, our Commercial Content Policy prohibits non-consensual sexual material and our Spotlight Terms require “you must have any necessary third-party rights including, without limitation, music copyrights and rights of publicity, for all content in your Snaps”. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to protect the privacy interests of our community. Users have the ability to report Snaps and the reporting menu includes options “They leaked / are threatening to leak my nudes”, “It’s an inappropriate Snap of me”, “It involves a child”, and “They are pretending to be me”. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not categorize or recommend content that violates users’ rights to private life. For example, we have terms, moderation and enforcement to prevent distribution of illegal / violating content. We also do not process sensitive category information. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. For example, Snap ensures that ads shown are in line with its Snap Advertising Policies which states that advertisements do not collect sensitive information or special category of data. We also ensure advertisers are not targeting specific individuals on our platform and that users do not feel like their privacy is being compromised by our advertising. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we have specific prevalence testing and transparency reports for sexual content and intrusion of privacy. We also monitor privacy-related inquiries as detailed above. 116 CONFIDENTIAL Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to sexual content and Teen safety which may impact users’ right to private life. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups e.g CIPL, FPF. Our content moderation policies provide de facto content moderation crisis protocol. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. Our Privacy Center offers a suite of information on our products, users’ choices to safeguard their privacy and how to contact us. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures in place for Teens. For example, Teens cannot create public profiles and if they post to Spotlight or Snap Maps their profile details are anonymized as an extra precaution. Our reporting menu also includes the option to report “It involves a child”. Our Family Center includes resources and guidance for Teens and their parents or trusted adults. Our new parents site provides additional guidance for parents and carers on risks and support.41 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Yes, Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. 41 https://parents.snapchat.com. 117 CONFIDENTIAL Conclusion Snap considers the overall risk to be within the Level 2 category given the stakes and theseverity of threats to privacy life, despite low prevalence. However, privacy is the first of Snap’sfour core platform governance values. We have robust protections in place, including clear termsand moderation. Snap enforces against these content violations robustly. We also mitigate risksthrough intentional product design choices and collaborate with experts, think tanks andresearchers on human rights, privacy and online safety to inform our approach. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures to address risks relating to the right toprivate life. There is no change in this conclusion from our 2023 Report. 4.2.4 Right to data protection We understand well the importance of ensuring that personal data is collected, processed orsecured appropriately. Depending on how and the extent to which Snapchatters use ourplatforms, significant volumes of the content published on Snapchat’s in-scope services, includingon Spotlight / Discover and Snap Map, is user generated images and videos. Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). Snapchat handles a significant volume of personal data relating to individuals in the EuropeanUnion. Depending on how and the extent to which Snapchatters use Snapchat, this could belimited to basic account information or it could extend, for example, to published images andvideos and metadata about the Snapchatter’s interaction with such content. Significant volumesof the content published on Snapchat’s in-scope services is user generated images and videoswhich might be related to individual Snapchatter creators and/or others. It is therefore more likelythan not that Snapchat’s in-scope services could cause an impact on an individual’s dataprotection rights if such personal data is not collected, processed or secured appropriately. We monitor the number and nature of Privacy and Data Protection requests we receive. In our2023 Report, we explained that: ● Our prevalence testing showed that “invasion of privacy” made up an Extremely Low PVPin August 2023 (see our Prevalence Testing chapter), Snap also receives low numbers ofprivacy-related queries We have been pleased to observe that this situation has continued: 118 CONFIDENTIAL ● Our prevalence testing showed that “invasion of privacy” had seen a further, substantialfall in prevalence. It is now at a very low level. However, despite low numbers of privacy related queries from our users, given the significantvolume of personal data being processed by Snap in relation to Snapchat, we continue to assessthis risk falls within our Low Likelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that if an online platform were to undermine the right to data protection, this could fallwithin our ‘severe harm’ category. Overall potential risk prioritization Considering the extent of the personal data being processed by Snapchat and our assessment ofthe risk of severe harm from negative effects on the right to data protection within the EuropeanUnion, we have assessed this to be a Level 1 overall potential risk prioritization, notwithstandingthe low incident of privacy related queries from recipients for Snapchat’s in-scope services. Thereis no change in this assessment from our 2023 Report. Snap's Mitigations Privacy is central to Snapchat’s values. When we first created Snapchat, we decided to build a platform with strong privacy principles, pioneering data minimization and messages that delete by default. We believe that visual communication and messages that delete by default give youngpeople the opportunity to express themselves without the pressures of public metrics andpermanence. Online platforms may have normalized having a permanent record of conversationsonline, but in real life, friends don’t break out their tape recorder to document every singleconversation for public consumption or permanent retention. This makes Snapchat feel less like apermanent record and more like a conversation with friends—allowing people to expressthemselves in the same way they would if they were just hanging out at a park with their friends. We put significant thought and consideration to ensure those principles are reflected into thearchitecture of our platform, and into the design and implementation of our products, policies, and enforcement actions. Since Snapchat’s inception, we have embraced a privacy and safety by-design approach and decided that our platform architecture and product choices should play a major role in risk-mitigation. They can be found in our privacy and safety by design principles. We have a dedicated cross-functional group responsible for compliance with these principles.This group brings together Legal, Policy, Engineering and Product. Material product changesrelating to Snapchat are reviewed by Legal and specialist engineering teams, as well as relevantmembers of the cross-functional group. We use our Safety and Privacy by Design principles to 119 CONFIDENTIAL help mitigate risks to Teens. We maintain Data Protection Impact Assessments (DPIAs) of ourprocessing of personal data to ensure we are confident this will not result in a high risk to therights and freedoms of individuals. Sections 5 and 6 contain many examples of how we designed Snap’s inscope services withprivacy and safety principles in mind. As a result of the measures that Snap takes to protect personal data and provide users withactionable tools and transparent information, Snap continues to receive low numbers ofprivacy-related queries. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat is a platform with strong privacy principles. These principles are reflected into the architecture of our platform. Product changes are subject to privacy by design reviews and we maintain data protection impact assessments. For example, our Lenses only require object detection rather than facial identification. Lenses can tell what is or isn’t a face, they do not identify specific faces, limiting data processing for the use of Lenses. Snap does not use any data collected by Lenses to customize the content that the user sees in Spotlight or Discover, nor is any data collected for advertising purposes. Besides, voice data collection of Snapchatters in the EU is off by default; it is only used to provide the service. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our Privacy Center provides a suite of policies, including our Privacy Policy and they are enforced. In our Content Guidelines for Recommendation Eligibility we inform creators “We inform these 120 CONFIDENTIAL standards with proactive moderation using technology and human review” and “you must have any necessary third-party rights including, without limitation, music copyrights and rights of publicity, for all content in your Snaps” This prevents any risk that users may not be aware that their content submitted to Spotlight is subject to automated and human review, and prohibits creators from depicting individuals in content without necessary rights. In our Snap Spotlight Submission and Revenue Terms we state “You understand that Snaps you submit to Spotlight are Public Content and may be visible to all Snapchat users, as well as non-Snapchat users on other services and websites”. This prevents the risk of creators being unaware that their Stories submitted to Spotlight become public and informs users that their content may be saved off Snapchat. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove content that violates users’ right to data protection. For example, on Discover, Media Partners are proactively moderated and only a small pool of Snapchatters are shown in Discover (“Snap Stars” or “Popular Users”. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not categorize or recommend content that violates users’ right to data protection. For example, users can opt out from personalized recommendations based on inferred interest and we do not process sensitive category information. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, our Advertising Systems has a suite of protections including: ● No microtargeting ● We offer controls to turn off most personalized ads. Users can learn more about their choices here How to Adjust My 121 CONFIDENTIAL Advertising \& Interest Preferences on Snapchat. ● We ensure that sensitive data is not being used for ad targeting ● We continue to trial evolving privacy enhancing technologies, such as third party data clean rooms, to provide advertisers with options to further minimize the privacy impact of Snap ad services. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we consult with experts and our community, and we also monitor and respond to privacy-related inquiries. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. No, we don’t cooperate with trusted flaggers in relation to data protection violations. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups e.g. CIPL, FPF. We also have a well-established protocol to deal with privacy incidents, as well as a Security Incident Response Policy. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on privacy protection in our Privacy Center. For example, we explain to users How We Rank Content in Discover, How We Rank Content on Spotlight – Snapchat Support and Snapchat Ads Privacy \& Transparency. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit disclosure of Teens’ data and we avoid nudge techniques to encourage Teens to change their privacy settings and select less privacy-enhancing choices. We offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and 122 CONFIDENTIAL support.42 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. General content authenticity measures Conclusion Snap considers the likelihood and the serious nature of the impacts on the right of dataprotection within the European Union to fall within our Level 3 overall potential risk of Snapchat’sin-scope services. Depending on how and the extent to which Snapchatters use these platforms,significant volumes of the content published on Snapchat’s in-scope services is user generatedimages and videos. It is therefore more likely than not that Snapchat’s in-scope services couldnegatively affect an individual’s data protection rights if such personal data is not collected, processed or secured appropriately, which is why Snap enforces its privacy principles robustly. Privacy is central to Snapchat’s values. We put significant thought and consideration into our privacy principles and those principles are reflected into the architecture of our platform. We have a cross-functional group responsible for compliance with our privacy and safety by design principles, we review product changes for impact to data protection rights and we maintain Data Protection Impact Assessments of our processing of personal data where appropriate to ensurewe are confident this will not result in a high risk to the rights and freedoms of individuals. Wereceive a low level of data protection queries as a result of the robust protections in place. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures against the risk of negative effects on dataprotection rights. There is no change in this conclusion from our 2023 Report. 4.2.5 Right to non-discrimination and freedom of religion We understand well that online platforms can be used to spread content that contains orpromotes discrimination for example by using discriminatory characteristics for targeting ads,biased algorithms used for recommender systems and content moderation, the spread ofdiscriminatory content, facilitating online harassment, disproportionately reporting accounts ofindividuals from marginalized (religious) communities based on user reports, etc. This risk poses a 42 https://parents.snapchat.com. 123 CONFIDENTIAL serious threat to the rights of EU citizens who are already vulnerable to abuse and haveencountered discrimination and marginalization historically. Without mitigations, content thatundermines the right to non-discrimination and freedom of religion could conceivably appear inany of Snapchat’s in-scope services displaying user generated content, including information invideos featured on Spotlight / Discover and Snap Map. Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). We have assessed the relative likelihood of Snapchat’s inscope services disseminating contentrelating to matters that may undermine the right to non-discrimination and freedom of religion,based on Policy Violating Prevalence (PVP) via random sampling and our transparency reportdata in Section 4.1 of this Report, as follows: Category Relative likelihood of risk occurring on Snapchat 4.1.2 Dissemination of illegal hate speech Extremely Low Likelihood 4.1.4 Dissemination of terrorist content Extremely Low Likelihood In addition, although we identified in Section 4.2.4 we placed risks to the right to data protectionin the highest relative likelihood category, we have noted overall the number of privacy and dataprotection related queries we receive are very low. When focused on algorithmic bias specifically,we have not received any material volume of queries, which is not surprising as we noted inSection 4.2.2 (Right to freedom of expression and information) that Snap’s inscope services donot generally provide a platform for political public content. We therefore continue to assess that the relative likelihood that the in-scope services ofSnapchat would have an actual or foreseeable negative effects of the right to human dignity fallswithin our Extremely Low Likelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that if an online platform were to undermine the right to non-discrimination andfreedom of religion, this could fall within our ‘serious harm’ category. 124 CONFIDENTIAL In addition, we have assessed the severity of harm caused by categories of content that mayundermines the right to non-discrimination and freedom of religion in Section 4.1 of this Report, asfollows: Category Relative likelihood of risk occurring on Snapchat 4.1.2 Dissemination of illegal hate speech Significant harm industry wide 4.1.4 Dissemination of terrorist content Serious harm industry wide Overall potential risk prioritization Snap would consider this risk to the right to non-discrimination and freedom of religion to fallwithin the Level 3 category overall. Although we consider the risk to fall within our serious harmcategory, there is a relatively low prevalence of hate speech, terrorism and bias concerns on ouronline platform. There is no change in this assessment from our 2023 Report. Snap's Mitigations To help ensure our policies against hate speech are enforced responsibly, our teams consult theexpertise and work of civil society organizations, like Access Now, human rights experts, lawenforcement agencies, NGOs, and safety advocates. We are constantly learning, and willcalibrate wherever necessary to ensure that our products and policies function to keepSnapchatters safe. Practically, our in-app reporting tool allows users to directly report hateful content or activitiesthat support terrorism or violent extremism. On our high-reach surfaces, like Spotlight andDiscover, we take a proactive approach to moderating any content that may violate these rules.When hateful content is reported, our teams will remove any violating content and users whoengage in repeated or egregious violations will have their account access locked. As anadditional measure, we encourage Snapchatters to block any users who make them feel unsafeor uncomfortable. For publicly available content on Spotlight, Discover and Snap Maps: ● We survey a subset of our users quarterly to understand whether they find their timespent on our experience entertaining and satisfactory. We use this to track whether ourproduct changes are improving viewers' overall perception of the app. ● We provide a diversity of perspectives. We have multiple programs to foster a more diverse content community and surface different perspectives (e.g. Black accelerator program.) ● We ensure there is always a large mix of content from creators from viewers’ homecountry and content in the language in which they have set their device. 125 CONFIDENTIAL ● We add diversity to every viewer’s feed in terms of the account they see, and thecategories of content we surface to them. This prevents users from entering an echochamber or filter bubble of seeing the same content repeatedly. We use machine learningto understand content categories and diversify it. Modifying facial features or overlaying cultural elements in Snapchat's Lenses may reinforcediscriminatory ideas based on appearance or ethnicity and promote harmful imagery. Also,Lenses incorporating cultural symbols or references might lack proper context and sensitivity. The Lens Studio Submission Guidelines, reiterated our Community Guidelines and spell out that the following categories of Lenses are prohibited: ● Content that demeans, defames, or promotes discrimination or violence on the basis ofany of the identities listed in our Community Guidelines ● Examples: slurs, stereotypes, hate symbols, the promotion of hateful conspiracy theories,the glorification of atrocities or historical hatemongers Snap designs every Lens with race, gender, ethnicity and cultural norms in mind. Snap leveragesits ever-growing divers training datasets, as well as feedback from community members. If a Lensdoes not resonate with our community, as expressed through a high ratio of user reports, we takethat feedback into consideration and will re-review the Lens with a goal to leave as-is, modify, orremove. If a Lens is appropriate, but could theoretically be misused by someone, that alone is notsufficient to reject a Lens. Snap considers current and historical global events when releasing aLens, and delays or denies amplification to Lenses that may be deemed insensitive due tobroader social occurrences throughout the world. Lenses should not change a user’s skin tone tomimic a different ethnicity or race. Snap does not modify facial or other features in a way thatevoke racial, ethnic, cultural or religious stereotypes or stigmatized disabilities. Snap presentsreligious and cultural iconography in a respectful manner, with feedback solicited from internaland external subject matter experts. This means Snap is especially thoughtful around holiday orevent-based content, including the geography in which a Lens will launch. Also, Snap ensuresthat a Lens is not deceptive. Snap uses signifiers and watermarks where there may be questionsof creative authenticity. Snap tests Lenses on photos/videos of and in real life settings withdiverse groups of people to accurately enforce our policies. As reported in our 2024 Diversity Annual Report, we know DEI is critical for long-term growth - whether it’s Snapchatters demanding products to meet their diverse needs, or the desire to reachnew and different markets. It also highlights in particular two new initiatives since our 2023Report that showcase examples of how empathy can inspire new perspectives and tangiblebusiness impact. ● Snap Out Loud \- Snapchat is a platform that celebrates authenticity. That’s why our teamcreated an AR experience to spotlight the different communities who share the LGBTQIA 126 CONFIDENTIAL umbrella. Led by SnapPride, this Snap Show educates Snapchatters about the meaning ofLGBTQIA, and celebrates the people who make up the community. Snapchatters werewelcomed into seven separate spaces, denoted by the letters of the acronym, to exploreeach world. 25 million unique users were reached across 11 countries and the lens wasshared one million times. ● 8 Mars 8 femmes \- In France, only 10% of statues in public spaces honor female figures.SnapWomen partnered with the Sales and AR Studio team in Paris to launch an ARactivation on International Women’s Day 2023 across 8 major cities. Called 8 Mars 8femmes, or 8th of March, 8 women, the activation featured female AR statues next tomale ones to celebrate great women in history who were never given appropriate creditfor their impact. These AR statues are permanently activated, and honor Josephine Baker,Olympe de Gouges, Manon Tardon, Hubertine Auclert, Simone Veil, Françoise deGraffigny, Élisabeth Vigée Le Brun, and Simone de Beauvoir. The initiative receivedwidespread media coverage in France. Snap’s AR Studio team was honored to win theglobal Drum Award for Marketing for their outstanding creativity. 127 CONFIDENTIAL Specific Mitigations. In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive moderation for hateful content or activities supporting terrorism or violent extremism. We also work with civil society organizations to ensure our policies are enforced responsibly. Product Inclusion helps us create equitable experiences by intentionally involving and 128 CONFIDENTIAL considering marginalized groups at critical moments throughout the product development process. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, Snap’s Terms and Community Guidelines prohibit Hate speech or content that demeans, defames, or promotes discrimination or violence on the basis of race, color, caste, ethnicity, national origin, religion, sexual orientation, gender identity, disability, or veteran status, immigration status, socio-economic status, age, weight, or pregnancy status. We strictly enforce these rules. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove hateful content or activities supporting terrorism or violent extremism. We provide in-app reporting for hateful content or activities supporting terrorism or violent extremism. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not categorize or recommend hateful content or activities supporting terrorism or violent extremism. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. In order to ensure we are not using discriminatory targeting models particularly when there is significant legal impact to the consumers, we offer special targeting models that do not include gender or age, which we require for advertisers who are advertising in the housing, credit or employment (HCE) spaces, so that discriminatory factors will not go into who sees these ads. We do not allow advertisers to build audiences for their ads based on their own data about our teenage users regardless of those user’s own ad settings (i.e. activity data from the advertisers own online properties and the advertiser’s own customer lists). Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we have specific prevalence testing and transparency reporting for hate speech, terrorist and violent extremist content. 129 CONFIDENTIAL Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to illegal hate speech, terrorist and violent extremist content. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups e.g. EU Internet Forum. We have signed onto the EU hate speech Code. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. We provide in-app reporting for hateful content or activities supporting terrorism or violent extremism. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures in place for Teens. For example, Teens cannot create public profiles and if they post to Spotlight or Snap Maps their profile details are anonymized as an extra precaution. Our reporting menu also includes the option to report “It involves a child”. We hope protections like these help protect Teens from hateful content. Our Family Center includes resources and guidance for Teens and their parents or trusted adults. Our new parents site provides additional guidance for parents and carers on risks and support.43 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to General content authenticity measures. However, Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. 43 https://parents.snapchat.com. 130 CONFIDENTIAL use functionality which enables recipients of the service to indicate such information. Conclusion Snap considers the overall risk to be within the Level 3 category taking account of the harm thatrisks to the right to non-discrimination and freedom of religion may cause and the low prevalencefor hate speech on the platform. In practice Snap has substantial protective measures in place.Snap works with civil society organizations, like Access Now, human rights experts, lawenforcement agencies, NGOs, and safety advocates to make sure we are calibrating wherevernecessary to ensure that our products and policies function to keep Snapchatters safe. Ourin-app reporting tool allows users to directly report hateful content or activities that supportterrorism or violent extremism. On our high-reach surfaces, like Spotlight and Discover, we take aproactive approach to moderating any content that may violate these rules. Further, our diversityand inclusion efforts continue to help us create equitable experiences and build inclusiveproducts. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures to protect users’ right to non-discriminationand freedom of religion. There is no change in this conclusion from our 2023 Report. 4.2.6 Children’s Rights We understand that online platforms can impact children’s and Teen’s rights. This is a risk we takeseriously as Snap’s priority is protecting the safety and wellbeing of our users whilst ensuringthey have a positive experience online. Privacy, safety and security are key values of thecompany and at the core of our value proposition to our users. The ‘rights of the child’ under the Charter44 comprises two elements that are relevant toSnapchat’s in-scope services: 1. Children have the right to such protection and care as is necessary for their well-being;and 2. Children have the right to express their views freely and have those views taken intoconsideration on matters which concern them in accordance with their age and maturity. In respect of element 1, we address the well-being of children when considering Category 4 ofthe DSA risks in particular parts of the negative effects on children and physical and mentalwellbeing elsewhere in this Section 4. This section therefore focuses on element 2 i.e. risks tochildren’s rights of expression. 44 Art 24, Charter of Fundamental Rights of the European Union (CFREU), url. 131 CONFIDENTIAL Likelihood As explained in Snapchat Community as part of our Introduction to this Report, Snapchat is usedby a wide demographic, with 18-24 year olds still making up the highest percentage of users ofSnapchat. Nevertheless, there is still a percentage of our users who are Teens (13-17)). Thereforewe still consider that children using Snapchat are just as likely to be exposed to freedom ofexpression issues identified in this Report as other members of the Snapchat Community asfollows: Risk Category Relative likelihood of risk occurring on Snapchat Relative likelihood of negative effect on children Right to freedom of expression Extremely Low Likelihood Extremely Low Likelihood As a result, we continue to conclude that the relative likelihood of a risk of negative effects onchildren and Teens for Snapchat’s in-scope falls within the Extremely Low Likelihood category. Severity We assessed the risk of harm from the right to freedom of expression to fall within our significantharm classification. However, we take the safety and wellbeing of the youngest members of ourcommunity very seriously and recognise that this group is particularly vulnerable and if aparticular risk materializes, there is an increased risk that the severity of the harm they suffer ishigher. For freedom of expression, we consider this as follows: Risk Category Harm classification industry wide Is the industry wide severity risk higher for children and Teens? Right to freedom of expression Significant harm industry wide Yes, Snap considers that it is vital that children and Teens are able to access online platforms and participate in lawful online debate and dialogue to learn, have their views heard and develop their own values and identities, regardless of their ability to pay. As a result, we have chosen to place the severity of harm arising from an issue that negativelyaffects children’s rights in our ‘severe’ category. 132 CONFIDENTIAL Overall potential risk Although the relative likelihood for the negative effects on children’s rights falls within ourExtremely Low Likelihood category, Snap considers the risk of harm to fall within the severestcategory. Consequently, Snap considers this to be a Level 1 overall potential risk for Snapchat’sin-scope services. There is no change in this assessment from our 2023 Report. As described in our risk methodology section, we assess overall potential risk on a case by casebasis and Snap reserves the option to deviate from the overall potential prioritization risk matrixwe use as a guide. This is one of the cases where we have chosen to deviate. Snap's Mitigations Highlights As explained in the freedom of expression and assembly part of this Section 4, we have put inplace a number of mitigations to ensure that all users, including Teens, have the right to expressviews freely, where appropriate: ● Our Terms clearly define certain topics which we prohibit, including false information thatthreatens public health (e.g. COVID-19 vaccinations), civic processes, or that denies tragicevents (like the Holocaust). We also have an explainer to help our community understandhow we handle harmful false or deceptive information. This provides clarity on the limitswe have when it comes to freedom of expression and assembly. ● Our platform is generally not a place for political or activist content. Such content is noteligible for promotion on Spotlight and user content on Discover is only from a smallnumber of popular, entertaining community creators and their content is moderated byhumans against our Content Guidelines. All Media Partners are vetted prior to beingpermitted to distribute their content broadly on Snapchat by a team of editors. ThesePartners include news organizations, which are subject to their own professional rules.Media Partners go through an editorial review of their content, a reputational search (toevaluate if a publisher has a history damaging press, legal actions, etc.), and compliancereview before they’re able to distribute content. As a result, we provide a balancedapproach to political and activist content on Snapchat that is designed to limit the sourcesof such information to professional media partners. ● Snapchat utilizes content moderation policies and systems to protect users’ rights tofreedom of expression and access to accurate information. As all of our user generatedcontent is moderated by a mix of automation and human moderation, we proactivelyremove content that does not meet our policies before being broadly distributed. In somecases, content against our policies may make it past moderation by mistake. In thosecases, we rely on Snapchatters to report the content for re-moderation. ● As explained when discussing the dissemination of content that infringes on intellectualproperty rights, Snap respects the doctrine of “fair use,” i.e., that there are certain 133 CONFIDENTIAL circumstances (such as news reporting, social commentary on issues of public interest,criticism, parody, or education) where excerpts of copyrighted material could bedistributed without permission from or payment to the copyright holder. This helpsreinforce the rights of freedom of expression and the freedom of assembly. We believe that it is also important that our business model supports the right for all users to useSnapchat, regardless of ability to pay, by paying for the cost of the service through balanced, andproportionate targeted advertising (as explained further in the data protection rights andadvertising systems section of this Report). This has been made more challenging by theobligation in the DSA to prohibit all forms of targeted advertising to Teens, even if balanced withreasonable, proportionate and effective mitigation measures in place. However, we continue tooffer Snapchat’s in-scope services to all, without charge, including Teens. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive moderation to protect Teens’ access to accurate information and provide an appropriate environment to meet, see new experiences and express themselves. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, terms provide clear rules to Teens on the boundaries of appropriate expression and prohibit harmful false or deceptive information. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making Yes, specific proactive and reactive moderation procedures that are fairly applied to protect users’ rights to freedom of expression and access to accurate information. On Snap Map, our editorial oversight protection for content showing up on Snap Map tries to strike the right balance between the need to preserve the Teen’s safety versus the free flow of information and expression. Examples of this include in February of last year when Russia moved into Ukraine, Snap Map 134 CONFIDENTIAL processes and dedicated resources for content moderation. developed tooling that allowed us to block all of Ukraine from creating content. This was in response to concerns that Russia was using it for their own strategic purposes (propaganda and tracking the movement of Ukrainians). Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, the pool of content recommended by our algorithmic systems does not generally include political or other important societal matters regardless of where they fall on the political spectrum. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we consult with our safety advisory board to ensure Snapchat is set up appropriately for Teens and monitor community reports for issues relating to freedom of expression. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. No, we do not work with trusted flaggers for Teen’s rights to freedom of expression specifically, however we are working with trusted flaggers on children’s safety in general. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, our Crisis Protocols balance Teen’s rights to freedom of expression with access to accurate information. We have recently exercised these protocols successfully during the French riots in June 2023. Note, we are actively working to support efforts to agree an EU Age appropriate design code to protect children’s rights. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms (including harmful false or deceptive information), moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping Yes, we have protective measures to allow Teens to express themself without the pressures of friends lists, comments and likes. We have community, ad and content guidelines that are specific to teens. We also offer Family Center; we make available robust 135 CONFIDENTIAL minors signal abuse or obtain support, as appropriate. reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.45 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Yes, for example we display an icon in some Lenses that manipulate an image of a Snapchat to make them look younger. Conclusion Snap considers children’s rights to be a lower likelihood risk but one that has a risk for severeharm industry wide, without appropriate mitigations. As a result we treat this as one of our highestpriority risks, with a Level 1 Risk Prioritization. Snap is designed to fairly apply rules on contentpublication and provide an appropriate environment for Teens to exercise expression andassembly on Snapchat’s in-scope services (and Snapchat as a whole). As explained in theFreedom of Expression and Protection of Minors section of the Report, this includes adapting oursystems to limit the access of Teen accounts to higher risk features and content, like publicprofiles and sexually suggestive content, as well providing Teens and Families with accessibleguidance and tools for the use of Snapchat and ensuring our Terms, Moderation andEnforcement also operate fairly. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures to protect against negative effects onchildren’s rights. In addition, we are actively participating in efforts to develop an EU wideAADC to assess if further industry measures are needed. There is no change in thisconclusion from our 2023 Report. 4.2.7 Right to consumer protection We understand that without mitigations online platforms can be used to spread content thatcontains false or misleading information that can harm consumers. This risk to consumer 45 https://parents.snapchat.com. 136 CONFIDENTIAL protection rights poses a serious threat to the rights of EU citizens who may be vulnerable todeception or invasion of privacy. Likelihood Snap has implemented safeguards, both through product design and policy enforcement, toeffectively diminish the likelihood that consumer protection rights are violated on the platform.We monitor the number and nature of Privacy and Data Protection requests we receive. In our2023 Report, we explained that: ● Our prevalence testing showed that “invasion of privacy” made up an Extremely Low PVPin August 2023, a decrease from April 2022. ● Snap also receives low numbers of privacy-related queries. We have been pleased to observe that this situation has continued: ● Our prevalence testing showed that “invasion of privacy” had seen a further, substantialfall. It is now at a very low level. ● Overall we have seen a slight increase in line with expectations and we continue toreceive low numbers of privacy-related queries. We monitor the number of community support requests we receive relating to the EuropeanUnion. Note, these figures (and the privacy figures above) concern the requests which we reviewmanually and excludes automated responses. We have observed a steady decrease in EUconsumer support requests. Finally, we also monitor for significant changes in our ad review processes. Given these safeguards we have in place, and the low, decreasing level of consumer queries andconsistent levels of ad rejections for our main fraud category, Snap considers that the risk to theright of consumer protection falls within the Extremely Low Likelihood. Severity Severity Snap has assessed information published by governments and other third party sources andconsiders that if an online platform were to undermine the right to non-discrimination andfreedom of religion, this could fall within our ‘serious harm’ category. As a result Snap treats risk of harm from a negative impact on the right to consumer protection as significant due to the potential harm it can cause to users. Overall potential risk prioritization Given we have assessed the potential for negative impacts to consumer protection rights to fallwithin the lowest likelihood category and to have a risk of significant harm, we consider this riskto fall within our Level 3 potential risk prioritization category, given the Extremely Low 137 CONFIDENTIAL Likelihood and significant harm categorization. There is no change in this assessment from our2023 Report. Snap's Mitigations Highlights To mitigate these risks, Snap takes a multipronged approach. Snap has invested considerableresources in developing and enforcing advertising policies that safeguard consumer protectionrights. We have robust ad policies to prevent inappropriate and illegal advertising on ourplatform, and we use a combination of automated and human review to prevent ads that violateour policies or the law from appearing on Snapchat. This also includes ensuring inappropriateads are not targeted at Teens. Additionally, all ads can be flagged by Snapchatters in the app asbeing inappropriate along with the reason for the violation. Separately, to ensure users know when content is commercial in nature, we automatically placean “Ad” marker on all paid ads that run on Snapchat. Our Commercial Content Policy requires allorganic content posted by influencers to be marked appropriately and we now offer a “PaidPartnership” tag tool that influencers and users may use when they post commercial content tohelp them comply with this policy and their legal obligations. To address potential risks with targeted advertisements, and to ensure advertisers are notmanipulating small audiences with micro-targeted campaigns, most of the ads on Snapchat,including all political ads, require a specific minimum audience of Snapchatters to be targeted.We also offer special targeting models that do not include gender or age for advertisers who areadvertising in the housing, credit or employment (HCE) spaces and are subject to specific legalrequirements relating to those ads. Lastly, to ensure that users have choice about use of theirpersonal data for targeting ads, we allow users to control the data that’s used to determine theads they see. In the EU, we offer controls to turn off most personalization of ads and for otherregions users can restrict our use of third party data and being included in advertiser suppliedaudience matches for ads targeting. Our Community Guidelines prohibit spreading false information that causes harm or is malicious,impersonation, i.e., attempting to deceive people about who you are, and disallow spam andother deceptive practices. Our Commercial Content Policy also disallows false or misleadingcontent, including deceptive claims, offers, functionality, or business practices, promotion offraudulent goods or services, products or services with false celebrity testimonials or usage,deceptive financial products, and other similar content. Through these mitigations, Snap hasbeen able to effectively uphold users’ consumer protection rights. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of the 138 CONFIDENTIAL DSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive moderation for ads that violate our policies, or false or deceptive content. Snap places a strong emphasis on its adherence to Article 25 DSA concerning dark patterns. Consequently, this constitutes a strategic mitigation measure aimed at mitigating the potential impediment to consumer protection. Snap is committed to ongoing monitoring of this aspect to ensure continued compliance and effectiveness. We also require a specific minimum audience of Snapchatters to prevent advertisers from manipulating small audiences. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, our advertising policies safeguard consumer protection and they are strictly enforced. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove ads that violate our policies, false or deceptive content. We also don’t allow user-generated political content from being promoted on Spotlight. We take these measures in order to circumvent the spread of harmful and false content. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not categorize or recommend false or deceptive content. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting Yes, our Advertising Systems are set up to safeguard consumer protection. For example, we automatically place an “Ad” marker on all paid ads. 139 CONFIDENTIAL the presentation of advertisements in association with the service they provide. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we have specific prevalence testing and transparency report false information, impersonation, spam and other regulated goods. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to regulated goods. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups e.g. CIPL, FPF. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we have an ad marker on all ads and provide transparency on our privacy practices including ads on our Privacy Center. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, for example we prevent inappropriate ads for Teens and advertising based on profiling. We make available robust reporting; and we offer Family Center and provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.46 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to Yes, Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. 46 https://parents.snapchat.com. 140 CONFIDENTIAL use functionality which enables recipients of the service to indicate such information. Conclusion Snap considers consumer protection risks to fall within our overall Level 3 potential riskprioritization category given the widespread availability of this false and misleading content onthe internet. In response it has put in place a range of mitigation measures. This includes, forexample, developing and enforcing advertising policies that safeguard consumer protectionrights. Our ad policies aim to prevent inappropriate and illegal advertising and our reviewprocesses were designed to enforce these policies. Because of safeguards in the product designand policy enforcement, to effectively diminish the likelihood that consumer protection rights areviolated on the platform, this risk falls within the lowest likelihood level. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate, and effective mitigation measures for the risk of negative effects on consumerprotection rights. There is no change in this conclusion from our 2023 Report. 4.2.8 Right to Property The property right that has a significant risk of being impacted by Snapchat’s in-scope services isthe right to intellectual property. This risk stems from the disclosure of such property contrary to the intellectual property rights of a natural or legal person. This is discussed above under Section 4.1.5 (Dissemination of content that infringes on intellectual property rights). In addition, we continue to consider there is a potential risk that individuals may harm someoneelse’s property while under pressure to create content that others find entertaining or humorous. This risk is discussed above under Section 4.1.9 (Dissemination of content encouraging or engaging in violent or dangerous behavior). 4.3 Category 3: Negative effect on public security (Article 34.1.c / DSA Recital 82) In this part of the Report, we explain the results of our assessment on actual or foreseeablenegative effects of Snapchat’s in-scope services on our democratic and electoral processes, civicdiscourse and public security as required by Article 34.1.c and Recital 82 of the Digital ServicesAct. We have assessed in particular negative effects on democratic processes, civic discourseand electoral processes, as well as public security. Category 3 - Negative effect on Public Security 141 CONFIDENTIAL Category Relative likelihood of risk occurring on Snapchat Harm classification industry wide Risk Prioritization Conclusions 4.3.1 Negative effect on Democratic and Electoral Processes Extremely Low Likelihood Severe harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective mitigations 4.3.2 Negative effect on Civil Discourse Extremely Low Likelihood Severe harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective mitigations 4.3.3 Negative effect on Public Security Extremely Low Likelihood Serious harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective mitigations 4.3.1 Negative Effect on Democratic and Electoral Processes The role of digital platforms in helping to shape information environments establishes asignificant nexus with democratic and electoral processes. As digital technologies such as Snapenable expression and access to information, the impact of these platforms on the free and fairexercise of political rights warrants careful attention, presenting risks to which Snap has longbeen vigilant. Likelihood As outlined in Section 1, a significant proportion of European citizens use Snapchat. As at 1August 2024, we have 92.4 million average monthly active recipients of our Snapchat app in theEuropean Union (EU), and significant recipient numbers in individual Member States. There is thepotential for public content on Snapchat to reach a sizable audience within the European Union(particularly within the 18-24 age bracket which accounts for the biggest share of our registeredaccounts). However, Snapchat’s platform architecture, combined with its commitment to responsible policyenforcement across our content surfaces, establishes unique safeguards against risks todemocracy. The steps Snap has taken to mitigate threats to democracy mean that likelihood issubstantially diminished. Independent reports of electoral interference on Snapchat are vanishingly rare. In connectionwith a major, high-profile election in 2022, we onboarded Snap to the Election Integrity 142 CONFIDENTIAL Partnership (EIP),47 a partnership among leading research centers and civil society organizationswho monitor online harms to democratic processes; as participants in the EIP threat escalationprogram, our teams received only one single incident report from the researchers monitoringrisks on Snapchat. We participated in the Commission’s stress test and multi-stakeholderroundtable dialogues ahead of the European Elections in 2024 and were able to successfully navigate the test exercises. As we reported in our election blog post on 24 June 2024, Snap saw a small uptick in reported activity, but did not receive or observe any material incidents or threats.Our moderation and reporting tools worked well, and none of the reported pieces of contentwere verified as misinformation on Snapchat. As highlighted in Section 4.1.10 (Dissemination of harmful false information), Snap’s own reportingmetrics confirm the limited occurrence of content harmful to democracy: ● Our prevalence testing has consistently shown a very low prevalence of ‘harmful falseinformation’. Our testing in July 2024 showed a further significant reduction inprevalence. ● Our most relevant transparency reporting category on this topic is “harmful falseinformation” which our policies define as including content that “undermines the integrityof civic processes.” False information continues to account for only 0.1% of the total of allcontent enforced on Snapchat. This figure remained steady throughout both halves of2022, as well as through the second half of 2023. We track “Impersonation”enforcements separately, and this similarly accounts for a very low percentage of ourenforcement actions (0.3% in the second half of 2023, compared with 0.2% in the secondhalf of 2022). Snap’s product design and policy practices continue to substantially reduce the likelihood ofnegative impacts on democracy. Our ongoing quantitative and qualitative analysis shows that therisk of potential negative impact on democratic and electoral processes on Snapchat falls into our Extremely Low likelihood category. Severity Snap has assessed information published by governments and other third party sources andconsiders that if content that has a negative effect on democratic and electoral processes wereto materialise on an online platform, this could fall within our ‘severe harm’ category. Accounting for the real-world examples illustrating the potential disruptive effects that digitalplatforms can have on democratic and electoral processes, we understand that a negative effecton democratic and electoral processes has a risk of severe harm if not properly mitigated. 47 Election Integrity Partnership (2020), url. 143 CONFIDENTIAL Overall potential risk prioritization Taking into account the real-world examples illustrating the potential disruptive effects ondemocracy, there is risk of a severe harm risk if not mitigated, As a result, we assess this risk tofall within our Level 3 overall potential risk prioritization category. There is no change in thisassessment from our 2023 Report. Snap's Mitigations Highlights Snapchat’s architecture and its commitment to responsible policy enforcement across ourcontent surfaces, establishes unique safeguards against risks to democracy. We understand wellthat online platforms may have a negative effect on the electoral processes and the exercise ofpolitical rights by amplifying digital disinformation or deceptive content relating to politicalmatters or processes. However, the steps Snap has taken to mitigate threats to democracy meanthat likelihood is substantially diminished. As we highlighted in our 2023 Report, Snap has for some time taken a multifaceted approach tomitigating negative impacts to democracy, including policy enforcement, product design, andexpert engagement. This approach aligns with Guideline for providers of VLOPs and VLOSEs onthe mitigation of systemic risks for electoral processes,48 as follows: Internal Processes As we have explained in this Report, and highlighted above, our assessment of risk prioritizationwith regards to negative effect on democratic and electoral processes and Snap’s measures tomitigate the risk, are guided by information on elements such as the presence and activity ofpolitical actors on the service and the number of Snapchatters in the EU and evidence regardingthe use of tactics, techniques and procedures for information manipulation. We also engage with experts in the information integrity and democracy and human rightscommunity to inform our approach. This includes collaborations and engagement with individualexperts (for example, former US Ambassador to the UN Human Rights Council Eileen Donahoe, inaddition to several others), as well as think tanks (such as the Atlantic Council’s Digital ForensicsResearch Lab) and research collaborations (such as the Election Integrity Partnership). Access to official information on the electoral process and Media literacy initiatives We have regularly partnered with governments around the world to inform Snapchatters aboutelections and invite them to go vote. We believe that civic engagement is one of the most powerful forms of self-expression and have previously worked with election authorities in France, Netherlands, and Sweden to raise awareness of elections and encourage participation. A recent 48 Guidelines for providers of VLOPs and VLOSEs on the mitigation of systemic risks for electoralprocesses, April 2024, url. 144 CONFIDENTIAL example was the ‘23 Dutch provincial election cycle. With the Dutch Ministry of the Interior, Snapdeveloped a lens where Snapchatters could place voting bins in their living room and answerquestions about the election with ‘true’ or ‘false’. By taking this quiz Snapchatters are increasingtheir knowledge about the elections and are reminded to go vote. The recent European elections saw even more first-time voters eligible to participate – followingthe decision by Belgium and Germany to join Austria, Malta and Greece in lowering the votingage to 16. Ahead of this year’s EU elections, we teamed up with the European Parliament on aspecial AR elections Lens that encourages people to get out and vote. During the election, weshared this Lens with all EU Snapchatters along with a message to remind them to vote and a linkto the Parliament’s election website. Snapchat partnered with the European Parliament and European Commission to promote their ‘Use your vote’ information campaign on elections, including a dedicated Lens, and their awareness campaign on the risks of disinformation and deceptive content. Snap sent a push notification to over 50 million users to urge them to vote in the elections. Although out of scopeof this Report, we note that we also further instructed My AI to avoid engaging on political topics.We are proud to have done our part to contribute to the highest observed turnout of the last 30years, with 51.08% of the 357 million eligible citizens participating in the election. Measures to provide users with more contextual information One key way in which we mitigate the risk is through intentional product design choices. Ourplatform does not, for example, provide an unvetted feed of algorithmically curated politicalinformation; we disallow all political content from Spotlight (our broadcast platform for User 145 CONFIDENTIAL Generated Content) and pre-moderate that surface to ensure that such political content is notdistributed.49 This safeguard ensures that Snap is not algorithmically promoting politicalstatements from unvetted sources, and generally reflects Spotlight’s function as an entertainmentplatform. (Consistent with our commitments to fundamental rights of expression and access toinformation, Snapchat provides other, non-algorithmically amplified spaces for users to expresstheir views and political observations, such as Chat and My Story; users can also seek access topolitical information from known publishers and creators whom Snap has on-boarded fordistribution on the Stories tab). Snap’s policies expressly prohibit content that undermines the integrity of elections and civicprocesses. Drawing from expert research from the Election Integrity Partnership,50 we orient thispolicy around four pillars of risk: ● Procedural interference: misinformation related to actual election or civic procedures,such as misrepresenting important dates and times or eligibility requirements forparticipation. ● Participation interference: content that includes intimidation to personal safety or spreadsrumors to deter participation in the electoral or civic process. ● Fraudulent or unlawful participation: content that encourages people to misrepresentthemselves to participate in the civic process or to illegally cast or destroy ballots. ● Delegitimization of civic processes: content aiming to delegitimize democratic institutionson the basis of false or misleading claims about election results, for example. We take steps to explain our policy approach to safeguarding democratic information environments through our Community Guidelines and periodic blog posts. As technologies have evolved, we have updated our policies to cover all content formats –whether created by a human or generated by artificial intelligence (including deep fakes). Inpreparation for the recent EU elections, we also: ● Signed up to the AI Elections Accord, alongside other technology firms, where we pledged to work collaboratively on tools to detect and limit the spread of AI generatedcontent which aims to deceive voters. ● Introduced contextual symbols to help our community understand when they areinteracting with Snap generated AI content. Snap does not allow Lenses that encourage a particular political perspective. In line with thisapproach, politically related Lenses are disabled in Discover. Snap also rejects Lenses that 50 Election Integrity Partnership, ‘Evaluating Platform Election-Related Speech Policies, October 2020, url. 49 For these purposes, “political content” means content related to political campaigns and elections,government activities, and/or viewpoints on issues of ongoing debate or controversy. This includes contentabout candidates or parties for public office, ballot measures or referendums, and political actioncommittees, as well as personal perspectives on candidate positions, government agencies/departmentsor the government as a whole. 146 CONFIDENTIAL perpetuate false information to elections (e.g. the wrong date). AR moderators are given strictguidance during elections to escalate misinformation. As a result, when it comes to the inscope content services of Snapchat, rather than takingmeasures to provide users with more context around disinformation and Foreign InformationManipulation and Interference (FIMI) content through labels and other indications, Snap’sapproach is to take steps to avoid recommending such content to a public audience in the firstplace (see below) and to remove such user generated content promptly when it is detected orreported. To the extent political content is distributed on in-scope services (i.e. political ads), Snaphas safeguards in place, which are detailed in the political advertising section below. Recommender Systems Content that is approved for broader audiences must comply with both our Community Guidelines and our Content Guidelines for Recommendation Eligibility. All Spotlight and non-professional user generated Discover content goes through both automoderation, and oftenhuman review, against these guidelines before it is eligible for recommendation to a wideaudience. As an additional safeguard, we monitor content that is achieving large-scale reach(and ensure a human reviews it) as a sort of “virality circuit breaker” and a means of checking thatour pre-moderation systems are working effectively. Any content that is reported will be reviewedagainst the guidelines again for compliance. Political Advertising Political content is only eligible for broadcast (aka algorithmic distribution) on Snapchat onsurfaces reserved for publishers or creators with whom Snap engages in partnership, or through advertising. Our political ad policies ensure that any political advertisements are subject to review and fact-checking before they are eligible for placement on Snapchat. We also preventadvertisers from manipulating small audiences with micro targeted campaigns, particularly forpolitical ads. We do so by requiring a specific minimum audience of Snapchatters to be targeted(including Dynamic Ads on Snapchat | Snapchat for Business). In 2021 Snap joined the Dutch Code of Conduct for political ads.51 Under this Code onlineplatforms agreed to acknowledge a responsibility in maintaining the integrity of elections andavoid dissemination of misleading content and messages inciting violence or hate speech ontheir platforms, committed to making key data on online political advertising available publiclyand help avoiding foreign interference in elections by banning political advertisements fromoutside the European Union, and putting in place a user-friendly response mechanism to answerquestions or solve problems related to the Dutch elections. 51 For more details url. 147 CONFIDENTIAL In preparation for the EU elections we also partnered with Logically Facts, a leading fact checking organisation and signatory of the EU Disinformation Code of Practice, to help fact check political ad statements across the EU. We do not require ads to label when advertisement includes generative AI content nor do requireadvertisers to disclose to us the tools they used to edit or create their ad creative. Instead, ourapproach is to subject all of our ads to a review process, and political ads are also subject to factchecking. Deceptive ads are rejected, irrespective of whether they use AI, photoshop, or otherdigital editing tools. Ads that are not deceptive, and otherwise comply with our Ad Policies, areapproved to run (and if they are a political ad, they must include a “paid for by” disclaimer and arecatalogued in Snap’s political ads library). Influencers Our commercial content policy requires all organic content posted by influencers to be marked appropriately. Commercial content that relates to the following is not permitted: ● Election-related content about candidates or parties for public office, ballot measures orreferendums, political action committees, and content that urges people to vote orregister to vote. ● Advocacy or issue content concerning issues or organisations that are the subject ofdebate on a local, national or global level, or of public importance. Examples include:content about abortion, immigration, the environment, education, discrimination and guns. We now offer a “Paid Partnership” tag tool that influencers and users may use when they postcommercial content to help them comply with this policy and their legal obligations. We makeclear that Snap restricts the paid promotion of political messaging to traditional ad formats. This isin order to be responsible to our community and to maintain transparency. Demonetisation of disinformation content The policies and other mitigations highlighted in this Section ensure that the placement ofadvertising does not provide financial incentives for the dissemination of disinformation and FIMIwith regards to electoral processes and hateful, (violent) extremist or radicalising content that caninfluence individuals in their electoral choices. Integrity of services As explained in this Section, we have appropriate procedures to ensure the timely and effectivedetection and disruption of manipulation of the service when this has been identified by them asa relevant systemic risk, taking into account the best available evidence. We explicitly prevent theuse of “any robot, spider, crawler, scraper or other automated means or interface to access”Snapchat; use of Snapchat “in any manner that could interfere with, disrupt, negatively affect orinhibit other users from fully enjoying” Snapchat and any “attempt to circumvent anycontent-filtering techniques we employ” on Snapchat. 148 CONFIDENTIAL When we determine that a user has violated our Terms, we may remove the offending content,terminate the relevant account, and/or notify law enforcement. We may also briefly limit thevisibility of content suspected of being illegal or otherwise violating our terms if needed to enabletime for human moderators to review and provide confirmation (known as “temporary softremoval”). See Section 5.5 for more information. Third party security and research The severity of these risks is reflected in the resourcing Snapchat has committed to partnershipsand collaborations with leading researchers and civil society organizations who are analyzingthreats to democratic information environments, including the Atlantic Council’s Digital ForensicsResearch Lab, the Center for a New American Security, the Stanford Cyber Policy Center andUniversity of Washington, and the Poynter Institute (which is also secretariat for the InternationalFact-Checking Network). Also reflecting its serious approach to this risk, Snapchat has agreed onvoluntary rules for the 2021 Dutch elections in a Code of Conduct, which governs transparencycommitments regarding online political advertisements during election campaigns (see below formore detail).52 Snap is also subject to audit under the DSA which includes a review of Snap’s compliance with itsrisk assessment and mitigation obligations. Fundamental rights As set out in Section 4.2 of this Report, in line with the requirements of the DSA, when assessingits risks and mitigations, Snap has paid due regard to: ● the protection of fundamental rights enshrined in the Charter of Fundamental Rights ofthe European Union, in particular the right to freedom of expression and of information;and ● the impact of measures to tackle illegal content such as public incitement to violence andhatred to the extent that such illegal content may inhibit or silence voices in thedemocratic debate, in particular those representing vulnerable groups or minorities. As explained above, our platform does not, for example, widely distribute an unvetted feed ofalgorithmically curated political information. Under our Content Guidelines for RecommendationEligibility, Political content is also not eligible for promotion in Spotlight, limiting the ability of anyuser to share political content with strangers on Snapchat, unless it’s from trusted news partnersand creators, and pre-moderate that surface to ensure that other such political content is notdistributed. Mitigation measures linked to generative AI 52 The Dutch Code of Conduct Transparency Online Political Advertisements, url. 149 CONFIDENTIAL Snap maintains robust policies––applicable to both the dissemination and the creation ofgenerative AI content––that function to mitigate risk and advance safety. Creation On-platform features for creating generative AI content are not part of Snap’s inscope servicesand are out of scope of this Report (save for certain commonplace ad creation tools).Nevertheless, outside of its DSA obligations, we note that Snap has developed several internalpolicies relating to generative AI. In particular, (1) Content and Product policies: We have developed a suite of policies that disallow thegeneration of harmful content (including deceptive political content). Our policy andmoderation teams work in partnership with engineering and data science colleagues toensure that our AI products are responsibly trained on these policy parameters. (2) Acceptable Use: We have similarly developed Acceptable Use Policies that prohibit theuse of our AI tools to attempt to generate violative content at the prompt-level. These aligned very closely with the rules for content dissemination, which are explained below. We have also introduced contextual symbols to help our community understand when they are interacting with Snap generated AI content. We have created a generative AI support page to explain our approach to Snapchatters and other stakeholders. Dissemination In the context of dissemination of content on Snapchat’s online platform, in scope of the DSA, weunderstand well that online platforms may have a negative effect on the electoral processes andthe exercise of political rights by amplifying digital disinformation or deceptive content relating topolitical matters or processes. Our Community Guidelines and Terms of Service set out the rules on what content is allowed on Snapchat. They are focused on preventing harm to Snapchatters and the broader communityfrom content and behaviour, whether or not caused by generative AI or any other form of IT tools(such as Photoshop). These rules apply to all content formats across our platform, includingcontent that is AI-generated. While the rules are agnostic to content format or creative tools, theCommunity Guidelines specifically note: “We implement safeguards designed to help keepgenerative AI content in line with our Community Guidelines, and we expect Snapchatters to useAI responsibly. We reserve the right to take appropriate enforcement action against accounts thatuse AI to violate our Community Guidelines, up to and including the possible termination of anaccount.” Our rules and internal enforcement guidance include clear provisions related to content risks forcivic discourse and electoral processes. In particular, our Community Guidelines prohibitspreading false information that causes harm or is malicious, such as denying the existence of 150 CONFIDENTIAL tragic events, unsubstantiated medical claims, undermining the integrity of civic processes, ormanipulating content for false or misleading purposes (whether through generative AI or throughdeceptive editing). As technologies have evolved, we have updated our policies to cover all content formats –whether created by a human or generated by artificial intelligence. Our Community Guidelines rules on false information refer to a more detailed Explainer that prohibits content that undermines the integrity of civic processes, or deep fake content or other media that ismanipulated for false or misleading purposes. The Community Guidelines further explain thatthese prohibitions extend to the following types of harmful content: ● Procedural interference: misinformation related to actual election or civic procedures,such as misrepresenting important dates and times or eligibility requirements forparticipation. ● Participation interference: content that includes intimidation to personal safety or spreadsrumours to deter participation in the electoral or civic process. ● Fraudulent or unlawful participation: content that encourages people to misrepresentthemselves to participate in the civic process or to illegally cast or destroy ballots. ● Delegitimization of civic processes: content aiming to delegitimize democratic institutionson the basis of false or misleading claims about election results, for example. Sharing such content will violate Snap’s Community Guidelines irrespective of whether it isAI-generated or user-generated, or whether it is generated on Snapchat or on another platform. Snap enforces these Community Guidelines fairly and consistently, using internal policies andguidelines, and applies outcomes that are commensurate with the severity of risk. Accounts thatwe determine are used to perpetrate serious, high-severity harms will immediately be disabled.For other violations of our Community Guidelines, Snap generally applies a three-partenforcement process: ● Step one: the violating content is removed. ● Step two: the Snapchatter receives a notification, indicating that they have violated ourCommunity Guidelines, that their content has been removed, and that repeated violationswill result in additional enforcement actions, including their account being disabled. ● Step three: our team records a strike against the Snapchatter’s account. A strike creates a record of violations by a particular Snapchatter. Every strike is accompanied bya notice to the Snapchatter; if a Snapchatter accrues too many strikes over a defined period oftime, their account will be disabled. This strike system ensures that Snap applies its policies consistently, and in a way that provideswarning and education to users who violate our Community Guidelines. The primary goal of ourpolicies is to ensure that everyone can enjoy using Snapchat in ways that reflect our values andmission; we have developed this enforcement framework to help support that goal at scale. 151 CONFIDENTIAL Snap has a suite of internal policies and guidelines to help our content review and trust andsafety teams apply the Community Guidelines to user generated content disseminated via ouronline platforms (such as Spotlight and Discover). They provide more granular information for ourcontent review teams. In preparation for the recent EU elections, we also signed up to the AI Elections Accord, alongside other technology firms, where we pledged to work collaboratively on tools to detectand limit the spread of AI generated content which aims to deceive voters. However, as notedabove, Snap’s product design and policy practices outlined above have been demonstrated to beeffective in mitigating the risks of deceptive political content, including content generated usingAI tools, from achieving meaningful scale on Snapchat and substantially reducing the likelihoodof negative impacts on democracy. As noted above, all of our ads are subject to review, andpolitical ads are also subject to fact checkingDeceptive ads are rejected, irrespective of whetherthey use AI, photoshop, or other digital editing tools. We continue to detect and monitor risks as outlined in Section 6 of the Risk Report (includingworking with our Safety Advisory Board on the intersection of safety and generative AItechnology) and adapt our mitigations accordingly. Cooperation with national authorities, independent experts and civil society organisations Snap has closely followed the negotiations on the EU AI Act and plans to continue to activelyengage and assess collaboration opportunities on the upcoming AI Act, as well as on the drawingof the related codes of practice for providers of general-purpose AI models and those regardingthe detection and labelling of artificially generated or manipulated content. More broadly, tackling risks stemming from generative AI requires (among others) broadindustry-wide technical solutions which have not been clearly identified so far. This is why Snap isactively engaging with its peers and industry experts in different fora to share best practices andadvance the technical debate. These partnerships, industry collaborations and efforts include: ● OpenAI: Although My AI is out of scope of this Report, the fact that My AI is powered byOpenAI’s ChatGPT, has led to a good working partnership with OpenAI. This allows thecompanies to share best practices, including with respect to content moderation. ● Tech Coalition / Working Groups on Generative AI: Snap is a member of the TechCoalition’s Working Group on Generative AI Content, and a member of the GenAI BriefingSubgroup. The Working Group on Generative AI Content meets regularly to facilitatedialogue and information- and idea-sharing around mitigating content-level generative AIrisks. The GenAI Briefing Subgroup meets periodically to plan expert briefings for TechCoalition members on topics related to Generative AI risks; such briefings have includedrepresentatives from government, law enforcement, civil society, and the researchcommunity. 152 CONFIDENTIAL ● Tech Accord to Combat Deceptive Use of AI in 2024 Elections: Snap was an initialsignatory to the Tech Accord to Combat Deceptive Use of AI in 2024 Elections. Thiscompact seeks to set expectations for how signatories will manage the risks arising fromdeceptive AI election content created through their publicly accessible, large-scaleplatforms or open foundational models, or distributed on their large-scale social orpublishing platforms in line with their own policies and practices as relevant to thecommitments in the accord. The Accord was announced at the Munich SecurityConference in February 2024. ● ITI AI Futures Initiative: Through its membership in the Information Technology Industry Council (ITI), Snap has participated alongside other private sector actors in the AI Futures Initiative. Led by technical and policy experts spanning the tech ecosystem, the Initiative is a forum through which participants are developing action-oriented recommendationsfor AI policy and working to address emerging questions around AI. Deliverables to date have included the issuance of Global AI Policy Recommendations to help guide governments around the world as to develop responsible regulatory approaches toAI-related issues. ● HackerOne - Red-Teaming Collaboration: Snap partnered with HackerOne on redteaming exercises to test the strict safeguards Snap has in place around AI. Together withHackerOne, we made significant developments in the methodology for AI safety redteaming that has led to a more effective approach to surfacing previously unknownproblems. We refer to the HackerOne blog for more details: https://www.hackerone.com/ai/safety-vs-security ● As an active member of the EU Internet Forum, Snap will support the upcoming dedicatedworking group on generative AI matters. ● We are also members of the Centre for Information Policy Leadership (CIPL) and theFuture of Privacy Forum (FPF) which work with industry stakeholders (like Snap), NGOsand government agencies in each region to advance a broad array of information topics.CIPL has been a leader in AI matters for many years through its dedicated AI Project andspecific Brazilian AI Project. Most recently, in Europe, CIPL has responded to the UKInformation Commissioner’s Office (ICO)’s consultations on Generative AI, and led variousforums on Accountable Governance of AI and AI Regulation in Brussels and the UK.Similary, FPF is working on AI Governance and other responsible Gen AI initiatives. Further, we actively engaged in the Commission’s public consultation on its proposed DSAElection guidelines, and similar consultations and queries raised by national DSCs. As shownabove, we have worked to update our risk assessment to take into account the recommendationsin those guidelines. Specific Mitigations In addition to the detailed highlights above, in the table below we indicate the specific measureswe have taken to mitigate this risk in respect of Snapchat’s in-scope services, using the definedlist of mitigations set out in Article 35 of the DSA. The primary purpose of the below table is to 153 CONFIDENTIAL indicate whether each specific mitigation category applies to this risk and the descriptions areillustrative rather than exhaustive. As many of our mitigations apply to all of the risks assessed inthis Report, to reduce duplication in this Report, each row in the tables provides a link in the lefthand column to a full summary of the specific mitigation in Section 5 of this Report whichexplains in more detail how each mitigation operates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, as outlined above our platform does not, for example, provide an unvetted feed of algorithmically curated political information; we disallow all political content from Spotlight (our broadcast platform for User Generated Content) and pre-moderate that surface to ensure that such political content is not distributed. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, we take steps to explain our policy approach to safeguarding democratic information environments through our Community Guidelines and periodic blog posts. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, we algorithmically prevent political content from being promoted on Spotlight. Snap does not allow Lenses that encourage a particular political perspective. In line with this approach, politically related Lenses are disabled in Discover. Snap also rejects Lenses that perpetuate false information to elections (e.g. the wrong date). AR moderators are given strict guidance during elections to escalate misinformation. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not promote political content on Spotlight. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, our political ad policies ensure that any political advertisements are subject to review and fact-checking before they are eligible for placement on Snapchat. We prevent advertisers from manipulating small audiences with micro targeted campaigns, particularly for political ads, by requiring a specific minimum audience of Snapchatters. Risk Detection and Management Yes, we review and monitor compliance with our internal terms, policies and procedures. 154 CONFIDENTIAL Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Our Trusted Flaggers are not typically focussed on this risk, but we welcome their input on this matter. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes. We engage with experts in the information integrity and democracy and human rights community to inform our approach. This includes collaborations and engagement with individual experts (such as former US Ambassador to the UN Human Rights Council Eileen Donahoe, global democracy scholar and Stanford Professor Larry Diamond, and several others), as well as think tanks (such as the Atlantic Council’s Digital Forensics Research Lab) and research collaborations (such as the Election Integrity Partnership). Additionally, we partner with governments around the world to inform Snapchatters about elections and invite them to go vote. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we proactively encourage our users to go to vote through interactive campaigns. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we limit exposure to political content to Teens, but do educate Teens with trusted new sources on current events and inform users how they can participate in a democratic society. We offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.53 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent We recognise the risk that generative AI could be used to generate harmful false misinformation, including deep fakes. Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on 53 https://parents.snapchat.com. 155 CONFIDENTIAL markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Snapchat’s inscope services. We also label political advertisements, and maintain a political ads library. Conclusion Snap considers the overall risk potential of negative impact on democratic or electoral processesto be in the Level 3 category, given severity of potential harm. However, as described above,Snap has put in place numerous specific mitigations, such as algorithmically preventing thepromotion of political content in Spotlight, enforcing political ad policies, and disallowing Lensesencouraging political perspectives. Further, the design and function of Snapchat is such that it isnot conducive for the widespread distribution of viral content and we provide robust in-appreporting, which further mitigates this harm. Snap recognizes the importance of democratic andelectoral processes, and in fact has created interactive campaigns to raise awareness andencourage users to vote. Our prevalence data and our continuing monitoring efforts cited aboveshow that our safeguards are effective at mitigating these risks on Snapchat. We have taken intoaccount the Commission’s recommendations set out in the Guideline for providers of VLOPs andVLOSEs on the mitigation of systemic risks for electoral processes when carrying out ourassessment. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate, and effective mitigation measures for the risk of negative effects ondemocratic and electoral processes. There is no change in this conclusion from our 2023Report. 4.3.2 Negative Effect on Civil Discourse We recognize that without adequate mitigations, digital content platforms like Snapchat cancontribute to negative effects on civil discourse. Across Snap’s various products, these riskscould include: ● The potential for personalized content and algorithmic biases lock users into echochambers, reinforcing existing beliefs and potentially leading to polarized communities,which hinders open dialogue. ● The risk of amplified dis- and misinformation negatively impacting public opinion onimportant civic issues. ● The possibility of amplification of extreme or sensational content to retain user attentionleading to heightened polarization and a hostile online environment. 156 CONFIDENTIAL Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). While it is rather difficult to classify the likelihood of such a comprehensive social issue, we canrevert to and deduce from the reporting data available to us. We are unaware of any third-partyreports identifying these risks on Snapchat. Our own reporting data suggests that policy violations related to harming civic discourse (i.e., our“harmful false information” and “hate speech” categories) are encountered rarely: ● In our 2023 Report, the prevalence of harmful false information was measured at anextremely low percentage, and represented just 0.1% of total content enforced in thesecond half of 2022. Our prevalence testing showed that illegal hate speech made up anextremely low percentage of Policy Violating Prevalence (PVP) on Snap in August of2023. ● We have subsequently observed a further substantial decrease in the prevalence ofcontent falling within these categories. Consequently, we continue to conclude that this risk still falls within our Extremely Low likelihoodcategory. Severity Snap has assessed information published by governments and other third party sources andconsiders that if content that has a negative effect on civil discourse were to materialise on anonline platform, this could fall within our ‘severe harm’ category. Overall potential risk prioritization Snap considers the dissemination of information with actual or foreseeable negative effects oncivic discourse to fall within our severe harm category. Given the apparent low prevalence,overall, this risk falls within our Level 3 potential risk prioritization category. There is no changein this assessment from our 2023 Report. Snap's Mitigations Highlights Snap’s policies prohibit the spread of “harmful false information,” which we define as falsecontent that may result in broadly distributed harm, or is malicious. Referencing our internal policy 157 CONFIDENTIAL guidance, Snap enforces content as harmful false information if both of the following elementsare present: ● Information is determined to be false ● The false information could cause “broadly distributed harm”. “Broadly distributed harm”refers to harms that undermine societal- or community-level safety or security; harms thatundermine public health; harms that undermine civic processes or the exercise of politicalrights; and harms that denigrate the memory or history of peoples and tragic events. In addition to our internal policies, Snap’s Community Guidelines also note that harmful falseinformation is prohibited and includes denying the existence of tragic events, unsubstantiatedmedical claims, or undermining the integrity of civic processes – all of which could contribute tonegative impacts on civic discourse. Snap policies also prohibit the use of hate speech, hate symbols, and/or content that valorizesthe perpetrators of, or denigrates the victims of, human atrocities such as genocide. We define hate speech as content that demeans, or promotes discrimination towards, anindividual or group of individuals on the basis of their race, color, caste, ethnicity, national origin,religion, sexual orientation, gender identity, disability, veteran status, immigration status,socio-economic status, age, weight, or pregnancy status. Our policies note that hate speech mayinclude references to people that are dehumanizing or that compare humans to animals on thebasis of these traits and categories. Hate speech also includes the valorization ofperpetrators––or the denigration of the victims––of hateful atrocities (e.g., genocide, apartheid,slavery, etc.), as well as the promotion of hate symbols. Under Snap’s policies, hate symbols include imagery that is intended to represent hatred ordiscrimination toward others, including those featured in the hate symbols database maintainedby the Anti-Defamation League (ADL).54 Snap establishes additional safeguards against risks to civic discourse on our surfaces that helpdistribute content algorithmically. All Spotlight and Discover content goes through bothautomoderation and human review before it is eligible for distribution to a wide audience.Content that is approved for broader audiences must comply with our Community Guidelines and our Content Guidelines for Recommendation Eligibility. Any content that is reported will be reviewed against these guidelines again for compliance. Snap has also made intentional product choices to mitigate risks to civic discourse; this includesthe absence of algorithmically promoted groups, which have been shown to contribute to echochambers and to be vectors for misinformation, with negative consequences for civil discourse.55 55The Verge, ‘Facebook will stop recommending health groups’, September 2020, url. 54 The ADL database is available at: url. 158 CONFIDENTIAL In addition, many of our surfaces are not ideal vehicles to cause risks to civil discourse. Forexample, unless saved to your Public Profile, Public Stories and Snaps on the Map are onlyavailable for a maximum of seven (7) days (and often much shorter), which limits their arc ofinfluence. Similarly, there is considerable technical expertise required to create a Lens, making ita difficult surface (compared to other third party platforms) to navigate for the purpose of broadlydistributed harm. To remain vigilant against threats to civil discourse, Snap engages with experts from across civilsociety and the research community who study information integrity and resilience to onlineharms. These engagements include consultations and collaborations with online safety experts(including those represented on Snap’s Safety Advisory Board), with organizations combatingonline hate (such as the Anti-Defamation League), and engagement with research organizations,including the Atlantic Council Digital Forensics Research Lab and the Digital Wellbeing. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, as outlined in the section on Democracy our platform does not, for example, provide an unvetted feed of algorithmically curated political information; we disallow all political content from Spotlight (our broadcast platform for User Generated Content) and pre-moderate that surface to ensure that such political content is not distributed. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, we take steps to explain our policy approach to safeguarding civil discourse information environments through our Community Guidelines and periodic blog posts. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, Yes, all Spotlight content goes through both automoderation and human review before it is eligible for distribution to a wide audience. Content that is approved for broader audiences must comply with our 159 CONFIDENTIAL where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Community Guidelines and our Content Guidelines for Recommendation Eligibility. Any content that is reported will be reviewed against these guidelines again for compliance. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not create echo chambers and ensure users are subject to different types of content and viewpoints. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, our political ad policies ensure that any political advertisements are subject to review and fact-checking before they are eligible for placement on Snapchat. We prevent advertisers from manipulating small audiences with micro targeted campaigns, particularly for political ads, by requiring a specific minimum audience. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we review compliance with our terms and processes. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Our Trusted Flaggers are not typically focussed on this risk, but we welcome their input on this matter. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes. We engage with experts in the information integrity and democracy and human rights community to inform our approach. This includes collaborations and engagement with individual experts (such as former US Ambassador to the UN Human Rights Council Eileen Donahoe, global democracy scholar and Stanford Professor Larry Diamond, and several others), as well as think tanks (such as the Atlantic Council’s Digital Forensics Research Lab) and research collaborations (such as the Election Integrity Partnership). Additionally, we partner with governments around the world to inform Snapchatters about elections and invite them to go vote. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we proactively encourage our users to go to vote through interactive campaigns. 160 CONFIDENTIAL Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we limit exposure to political content to Teens, but do educate Teens with trusted new sources on current events and inform users how they can participate in a democratic society. We offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.56 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. We recognise the risk that generative AI could be used to generate harmful false misinformation, including deep fakes. Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. We also label political advertisements, and maintain a political ads library. Conclusion Similar to our conclusion on negative impacts to democracy and elections, Snap considers theoverall risk potential of negative impact on civil discourse to be in the Level 3 category, given theseverity of potential harm of dis- and mis-information and online echo chambers which can createa hostile online environment. In response, Snap has put in a range of mitigation measures that, in most cases, overlap with themitigations for risks to democracy and elections. These mitigations include proactive contentmoderation, enforcement of our Community Guidelines and Terms, a restriction on politicalcontent which is a high risk area for dis- and mis-information, and engagement with outsideexperts and trusted flaggers. Snap also takes positive, rather than reactive or punitive mitigations,including encouraging Snapchatters to vote and participate in civil discourse, and audienceminimums to preempt ad microtargeting. We take facilitating and encouraging civil discourse very seriously, and view this matter asimportant to the value of Snapchat to our users. As such, we continue to invest in measures toprevent any content that negatively impacts civil discourse from reaching a broad audience onSnapchat, which may undermine our goal of allowing users to live in the moment and enjoy theworld around them. We also provide users with tools to report content and support resourcesonline and in-App, and we hold our advertisers to standards that prevent false, misleading, or 56 https://parents.snapchat.com. 161 CONFIDENTIAL micro targeted advertising. We specifically monitor the dissemination of harmful false information(Section 4.1.10) and dissemination of illegal hate speech (Section 4.1.2), which are consideredprimarily sources of negative effects on civil discourse. Both have been observed to have verylow prevalence on Snapchat’s inscope services such as Discover and Spotlight. As a result, we have concluded that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures against negative effects on civil discourse.There is no change in this conclusion from our 2023 Report. 4.3.3.Negative Effect on Public Security Without appropriate mitigations, we recognise that digital platforms may present risks to publicsecurity, particularly in the form of harmful, dangerous, or inciteful content; these risks maybecome compounded when such content may be amplified at great scale and distributed withhigh velocity. The design of Snap’s products and platform architecture scrupulously accounts forthese risks; accordingly, we’ve implemented a number of key safeguards that help to advanceboth the safety of Snapchatters and the interests of public security across our services. Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). To assess the likelihood of Snapchat’s inscope services having a negative effect on publicsecurity, we have reviewed the sources of data relating to the following illegal and otherwiseviolating content categories considered to have a particular impact on undermining publicsecurity: Category Relative likelihood of risk occurring on Snapchat 4.1.2 Dissemination of illegal hate speech Extremely Low Likelihood 4.1.4 Dissemination of terrorist content Extremely Low likelihood 4.19 Dissemination of content encouraging or engaging in violent or dangerous behavior Extremely Low Likelihood 4.1.10 Dissemination of harmful false misinformation Extremely Low Likelihood The low prevalence rate of these harms supports our continued assessment that the volume ofcontent presenting risks to public security is quite low on Snapchat, and, consequently, it isuncommon to encounter these harms on Snapchat. In terms of likelihood, this risk would fallwithin our Extremely Low likelihood category. 162 CONFIDENTIAL Severity Snap has assessed information published by governments and other third party sources andconsiders that if content that has a negative effect on public safety were to materialise on anonline platform, this could fall within our ‘severe harm’ category. To assess the likelihood of Snapchat’s inscope services having a negative effect on publicsecurity, we have reviewed the sources of data relating to the following illegal and otherwiseviolating content categories considered to have a particular impact on undermining publicsecurity: Category Relative likelihood of risk occurring on Snapchat 4.1.2 Dissemination of illegal hate speech Significant harm industry wide 4.1.4 Dissemination of terrorist content Serious harm industry wide 4.19 Dissemination of content encouraging or engaging in violent or dangerous behavior Significant harm industry wide 4.1.10 Dissemination of harmful false misinformation Significant harm industry wide As these range from significant to serious, and given the context outlined above showing seriousconsequences of a negative effect on public security, we assess that this category would fallwithin a ‘serious’ harm category. Overall potential risk prioritization Taking into account the real-world examples illustrating the potential disruptive effects on publicsecurity, this is a severe risk if not mitigated. However, we are encouraged––based on relevantprevalence data cited above––that our safeguards are substantially effective at mitigating theserisks on Snapchat. The combination of low prevalence and severe nature results in a Level 3 overall potential risk prioritization categorisation. There is no change in this assessment from our2023 Report. Snap's Mitigations Highlights We take several steps to ensure that we are addressing this risk across Snap’s products andservices, including enforcement of several relevant platform policies and internal crisis protocolsfor managing high-risk scenarios. 163 CONFIDENTIAL Snap’s policies include several prohibitions that are enforced vigorously and equitably to supportthe interests of public security. These policies include a prohibition against spreading harmfulfalse information. Internal policy guidance instructs that violations of these policies include risksto public security such as Snaps denying the holocaust or a school shooting, or informationobtained illegally that is being shared to embarrass the person from whom the information wasstolen. Snap’s policies also include prohibitions on content promoting terrorism or violent extremism, aswell as “content that attempts to incite, glorify, or depict real violence that results in personalinjury or death,” and “depictions of human violence, child abuse, animal abuse, or gore.” We may also consider off-platform behavior when assessing risks to public security. OurCommunity Guidelines state expressly that “Snap reserves the right to remove or restrict accountaccess for users whom we have reason to believe, in our sole discretion, pose a danger toothers, on or off of Snapchat. These include leaders of hate groups and terrorist organizations,individuals with a reputation for inciting violence, perpetrating severe harms against others, orbehavior that we believe poses a threat to human life.” Taken together, these several policy provisions provide a basis for appropriately actioning anycontent that poses an acute risk to public security. In addition, we have internal operational protocols for responding to public crises (see Section5.12). These protocols include the following steps: ● Our vendor teams carefully apply the Community Guidelines and Content Guidelines for Recommendation Eligibility to ensure the content is assessed appropriately against our rules (for example, routinely distinguishing between documenting violence and advocating for violence). ● When breaking news happens, such as ongoing violent protests, the vendor teamsconnect with our full-time content review team to summarize the kind of content they areencountering (e.g., violence, property damage, fires, expressions of criticism or supportfor various political positions), and summarize how they are currently actioning that typeof content against our existing guidelines. ● That summary list comes to our Platform Policy team for review. Almost all of the time,Policy’s answer is that they're actioning content correctly. (To cite a recent example, in thecase of French protests over the course of this summer, our team determined that existingpolicies and procedures were working as intended.) ● In the event that the Platform Policy team determines that the policies are not beingapplied appropriately, the team will expeditiously draft clarifying guidance for vendors andcontent review teams The draft guidance will be shared among relevant internal leadersfor review before being distributed to operational teams. 164 CONFIDENTIAL In addition, as noted in the Civil Discourse section, many of our surfaces are not ideal vehicles tocause risks to public security. For example, unless saved to your Public Profile and Public Storiesare only available for a maximum of seven (7) days (and often much shorter), which limits their arcof influence. Similarly, there is considerable technical expertise required to create a Lens, makingit a difficult surface (compared to other third party platforms) to navigate for the purpose ofbroadly distributed harm. Separately, we maintain tight internal protocols for escalating terrorist content or other imminentthreats to the appropriate legal or emergency authorities. In such cases, vendors and reviewteams are trained to preserve relevant information and immediately send a report to Snap’s LawEnforcement Operations team, who are professionally trained to appropriately engage with legaland emergency authorities. This approach reflects Snap’s deep commitment to public safety, and serves our community wellto reduce negative impacts to public security. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat is not an attractive platform for spreading content that may have a negative impact on public security, including harmful, dangerous, and inciteful content, in particular because it is difficult to reach a broad audience, and Snapchat has made conscious design decisions to restrict the ability for content to go viral, including not offering a reshare functionality and applying short retention to content. On surfaces where a broader audience can potentially be reached our proactive detection makes it difficult for content that may have a negative impact on public security to reach a large audience. Moreover, our content platform, Discover, features content from approved media publishers and content creators. Our entertainment platform for user-generated content, Spotlight, is proactively and a priori 165 CONFIDENTIAL moderated before content can reach a wide audience. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, terms prohibit content that may have a negative impact on public security, including harmful, dangerous, and inciteful content, and they are strictly enforced. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove content that may have a negative impact on public security, including harmful, dangerous, and inciteful content. Content on Spotlight undergoes rigorous moderation practices as reported in the Moderation section as well as under Risk Category 1\. All content on Spotlight is subject to human review pursuant to our Broadcast UGC Policies that are further described in Moderation. All Discover UGC content is moderated by humans, and we proactively remove content that doesn’t meet our policies before being broadly distributed. Furthermore, all Media Partners are vetted prior to being permitted to distribute their content broadly on Snapchat by a team of editors. Media Partners go through an editorial review of their content, a reputational search (to evaluate if a Media Partner has a history damaging press, legal actions, etc.), and compliance review before they’re able to distribute content. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not categorize or recommend content that may have a negative impact on public security, including harmful, dangerous, and inciteful content. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Yes, we have specific prevalence testing and transparency reporting for content that may have a 166 CONFIDENTIAL Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. negative impact on public security. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers, our trusted flaggers may also report content that may have a negative impact on public security, but this rarely happens because of the limited amount of this type of content on Snapchat. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.57 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. We recognise the risk that generative AI could be used to generate harmful false misinformation, including deep fakes. Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. We also label political advertisements, and maintain a political ads library. 57 https://parents.snapchat.com. 167 CONFIDENTIAL Conclusion Snap considers the negative impact to public security to have a Level 3 overall potential riskgiven the potential disruptive effects of content that can, among other things, harm, put in danger,and incite the public at large. That being said, Snap has put in place a range of mitigationmeasures to bring the likelihood of this risk from coming to fruition into the lowest category.These measures include our proactive content moderation which is designed to detect andprevent hateful, dangerous, and inciteful content from reaching a broad audience on Snapchat’sin-scope services. As noted in other sections, we continue to invest in measures that prevent thistype of content from reaching a broad audience on Snapchat, as well as provide our users withtools to report content to Snapchat and law enforcement, and support our community via onlineand in-app support tools. As a result, the volume of content presenting risks to public security islow on Snapchat. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate, and effective mitigation measures for the risk of negative effects on publicsecurity. There is no change in this conclusion from our 2023 Report. 4.4 Category 4: Negative Effects on Public Health (Article 34.1.d / DSA Recital 83) In this part of the Report, we explain the results of our assessment on actual or foreseeablenegative effects of Snapchat’s in-scope services on our public health as required by Article 34.1.dand Recital 83 of the Digital Services Act. We have assessed in particular negative effects onpublic health, gender-based violence, Teens, as well as serious negative consequences to aperson’s physical and mental well-being. We have considered risks relating to the design,functioning or use, including through manipulation such as by coordinated disinformationcampaigns related to public health, or from online interface design that may stimulate behavioraladdictions of recipients of the service. Category 4 - Negative Effects on Public Health Category Relative likelihood of risk occurring on Snapchat Harm classification industry wide Risk Prioritization Conclusion 4.4.1 Negative Effects on Public Health Extremely Low Likelihood Severe harm industry wide Level 3 Low Risk / Reasonable, proportionate and effective mitigations 4.4.2 Negative Effects on Extremely Low Likelihood Serious harm industry wide Level 2 Low Risk / Reasonable, 168 CONFIDENTIAL gender-based violence proportionate and effective mitigations 4.4.3 Negative effects on Children Varies Severe harm industry wide Level 1 Low Risk / Reasonable, proportionate and effective mitigations and Snap is participating in efforts to develop an EU wide guidance to assess whether further measures should be taken industry wide. 4.4.4 Serious Negative Consequences on physical and mental well-being Extremely Low Likelihood Severe harm industry wide Level 1 Low Risk / Reasonable, proportionate and effective mitigations 4.4.1 Negative Effects on Public Health We recognize that without adequate mitigations, digital content platforms like Snapchat couldcontribute to negative effects on public health. We believe the health and wellness of the publicand our users is paramount to our goal to be a platform of fun and freedom of expression. While we believe these risks to be probable in the absence of mitigations, we assess that Snap’smitigations appreciably reduce the likelihood of encountering these harms on our services. Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). With regards to coordinated disinformation campaigns related to public health, as well as anydissemination of content that promotes harmful/unhealthy behavior (e.g., eating disorders orother self-harm content), we are encouraged that available data suggests prevalence of contenton Snapchat related to these risks is quite low: ● In our 2023 Report, we observed that we had a low prevalence of health misinformationand other harmful false information to be extremely low and self-harm content isextremely low based prevalence testing. ● We are pleased to have subsequently observed a further fall in provenance of thiscontent, and the PVP rates are now 0.001 % for health misinformation and other harmfulfalse information and 0.014% for self-harm and suicide. For the potential negative effects on physical and mental well-being, see the relevant section. 169 CONFIDENTIAL Therefore, we continue to assess there is an extremely low likelihood of negative effects onpublic health arising from Snapchat’s in-scope services. Severity Snap has assessed information published by governments and other third party sources andconsiders that if content that has a negative effect on public health were to materialise on anonline platform, this could fall within our ‘severe harm’ category. Overall potential risk prioritization Snap assesses negative impacts to public health to present a systemic, severe risk that must beappropriately mitigated. Given the severe nature but extremely low prevalence, this wouldclassify as a Level 3 risk in terms of Snap’s overall potential risk prioritization matrix. There is nochange in this assessment from our 2023 Report. Snap's Mitigations Highlights Snap’s Community Guidelines prohibit the spread of harmful false information, expressly disallowing content that includes unsubstantiated medical claims. Our policies elaborate thatsuch prohibited content includes any content that, for example, recommends untested therapiesfor preventing the spread of Covid-19; or that features unfounded conspiracy theories aboutvaccines. Our Community Guidelines also prohibit “the glorification of self-harm, including the promotion of self-injury, suicide or eating disorders.” Our global approach to harm mitigation relies on teams,technologies, policies, and partnerships to help keep Snapchatters safe, healthy, and informed;however, content related to self-harm and suicide implicates unique sensitivities for which ourefforts must account. We therefore take a tailored approach to this category of content – one thatincludes: (1) supportive interventions, (2) features promoting a culture of support, and (3) aconsiderate approach to policy enforcement and self-harm content removal. Each of these will beexplained briefly in turn: 1. Supportive Interventions: In response to troubling search inquiries or content indicatingmental or emotional distress, our products and teams intervene to surface mental healthresources and support (either automatically, or at the discretion of Trust and Safetypersonnel). These resources are tailored to a user’s geographic region. 2. Culture of Support: Snapchat offers well-being features designed to educate andempower members of the Snapchat community to support friends who might be 170 CONFIDENTIAL struggling with their social or emotional well-being. These features include “Here for You” content Snap has developed with the intention of educating Snapchatters about theimportance of mental health, and ways to seek support. 3. Considerate Policy Enforcement: Especially given the risks of glorification, Snap’s policiesprohibit content that depicts suicide or self-harm; however, since Snapchat is used forcommunication with friends and family, it is important to us that our enforcement actionsdo not deprive users’ friends and family of important distress signals and an opportunityto intervene. Accordingly, we instruct reviewing agents that: ○ Reported depictions of suicide or self-harm that reflect an emergency situationshould be removed and possibly escalated to law enforcement or emergencyauthorities. ○ Content glorifying or inciting self-harm must be removed and is subject to anenforcement “strike.” ○ Depictions of self-harm or suicidal ideation that do not reflect an emergencysituation are permitted so that the community of people around this person canoffer help and support. To inform a responsible approach to mitigating these risks to public health, Snap regularlyengages with experts from across the field of online safety, health, and wellbeing. Our SafetyAdvisory Board includes several such experts (including, for example, Dr. Michael Rich,pediatrician, founder and director of the Digital Wellness Lab \& Clinic for Interactive Media andInternet Disorders, with affiliations at Boston Children's Hospital and Harvard Medical School).These experts have been consulted specifically on Snap’s approach to wellness and mitigatingrisks related to mental and emotional duress, eating disorders, and other forms of self harm. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive moderation to reduce the spread of harmful false information, including unsubstantiated medical claims, and the glorification of self-harm, including the promotion of self-injury, suicide or eating disorders. 171 CONFIDENTIAL Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, Our Terms prohibit the spread of harmful false information, including unsubstantiated medical claims, and the glorification of self-harm, including the promotion of self-injury, suicide or eating disorders. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove harmful false information, including unsubstantiated medical claims, and the glorification of self-harm, including the promotion of self-injury, suicide or eating disorders. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not categorize or recommend harmful false information or content that glorifies self-harm Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we have specific prevalence testing and transparency reporting for harmful false information and self-harm content. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to illegal content that harms public health. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups e.g. EUIF. We also coordinate with our Safety Advisory Board on issues related to public health, as it contains experts from the medical community. Transparency Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the 172 CONFIDENTIAL Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Annex), as well as how to and how to get help in our Safety Center. This includes specific new resources for sextortion in an effort to support those in distress. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. We seek to protect all users from these harms. We offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.58 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Yes, Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. This would include harmful false information, such as deep fakes. Conclusion Snap recognises that without adequate mitigations, digital content platforms like Snapchat cancontribute to negative effects to public health. In response, it has put in place a range ofmitigation measures. These include in particular our proactive content moderation, which isdesigned to detect and prevent content that may contribute to negative effects to public healthfrom reaching a broad audience on Snapchat. Given that we have observed a further substantialfall in prevalence rates for harmful false information and self-harm and suicide content, and theseare now at extremely low levels, we believe our mitigations have been effective. We continue toinvest in measures that prevent this type of content from reaching a broad audience onSnapchat, as well as provide tools to support our community via online and in-app support tools. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate, and effective mitigation measures for the risk of negative effects on publichealth. There is no change in this conclusion from our 2023 Report. 58 https://parents.snapchat.com. 173 CONFIDENTIAL 4.4.2 Negative Effects on gender-based violence We strongly oppose content that promotes gender-based violence. We recognise that withoutmitigations, a recipient of an online platform’s services could Promote content considered to begender-based violence. Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). Snap does not track gender-based violence as a specific, separate category in its prevalence andits transparency reports. However, this type of content is captured within the scope of broadercategories tracked by Snapchat including content relating to hate speech, harassment, harmfulfalse information, and violence and we have observed very low levels of prevalence of this typeof content on Snap’s inscope services: ● In our 2023 Report, we were encouraged by data suggesting the likelihood ofencountering such risks on Snapchat is within the lowest level. For example, data indicatethat the prevalence (PVP) of hate speech is extremely low ; for harmful false information;and all forms of harassment (including NCII, sexual harassment, and sextortion). ● We are pleased to have subsequently observed a significant fall in prevalence across allof these content types. From this, we continue to conclude that content promoting gender based violence falls within our Extremely Low Likelihoodcategory relative to other risks we have assessed. Severity Snap has assessed information published by governments and other third party sources andconsiders that if content that has a negative effect on gender-based violence were to materialiseon an online platform, this could fall within our ‘serious harm’ category. Overall potential risk prioritization Although the prevalence of content within the scope of this potential risk on Snapchat isconsidered to be at a lower level, due to the potential for serious harm to be caused by thiscontent, Snap considers this to be a Level 2 overall potential risk for Snapchat’s in-scopeservices. There is no change in this assessment from our 2023 Report. As described in our risk methodology section, we assess overall potential risk on a case by casebasis and Snap reserves the option to deviate from the overall potential prioritization risk matrixwe use as a guide. This is one of the cases where we have chosen to deviate. 174 CONFIDENTIAL Snap's Mitigations Highlights Snap takes a multifaceted approach to mitigating risks that may negatively impact gender-basedviolence. Our policies include several prohibitions against content that may contribute to suchrisks, including sextortion, sexual harassment, NCII, harmful false information (which may includegender-based disinformation campaigns), hate speech, and human trafficking. See details of themeasures we put in place to mitigate these risks in section 4.1 above. We also undertake intentional efforts to help all stakeholders understand these problems across the online community. As part of our Year Two Digital Well-Being study, we conducted a deeper drive into teens’ and young adults’ exposure to “sextortion” across platforms and services. Thetarget countries were Australia, France, Germany, India, the UK, and the U.S, which includes threeof the largest European countries, two of which are in the EU). We have continued this researchduring 2024 (“Year Three”), and also investigated teens’ and young adults’ attitudes andsentiments around reporting problematic content to platforms and services, authorities andothers. More information on this research can be found in Section 6.6. We believe our approach to these challenges reflects our commitment to responsibly mitigatingharms that may negatively impact gender-based violence. We received 67,152 reports of HateSpeech content in the EU in the second half of 2023, which led to enforcement action beingtaken against 8,894 unique content items and 15,007 accounts. This suggests users are able andwilling to report content and accounts as needed on Snapchat (note these numbers relate to thewhole of Snapchat, including private spaces that are out of scope of the DSA). We areencouraged by evidence that our approach has contributed to a low prevalence of CSEAI andhate speech content on the inscope services of Snapchat. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive moderation for illegal hate speech and violence, which includes gender-based violence. 175 CONFIDENTIAL We also take gender-based violence into consideration across the design of our services. For example, some Lenses can be used with Friends. Snap has removed tips to “Try with Friends” to some Lenses where there is a risk for bullying or harassment, including in relation to gender-based violence. In risky cases, Snap won’t encourage users to try a Lens with friends or Snap disables the Lens for being used with the rear camera (e.g. disabling this for the Pride Lens limits the ability to out someone else). These restrictions only applies to Lenses created by Snapchat. For Lenses submitted to Snapchat, we reject harmful Lenses to reduce the likelihood that they are distributed on Snapchat. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes, terms prohibit gender-based violence and they are strictly enforced. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Yes, specific proactive and reactive moderation procedures to prevent and remove violent content, and illegal hate speech, as further detailed in previous sections of this report. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not categorize or recommend violent or illegal hate speech content, which would include gender-based violence, as further outlined in the previous sections. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of Yes, as outlined in the previous sections on hate speech and violent content. 176 CONFIDENTIAL their activities in particular as regards detection of systemic risk. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to illegal hate speech. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups e.g. EUIF. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, we have protective measures to limit Teen contact with strangers; we offer Family Center; we make available robust reporting; and we provide guidance to parents on the web. Our new parents site provides additional guidance for parents and carers on risks and support.59 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Yes Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. This includes hate speech and violent content. Conclusion Similar to the related hate speech risk category, Snap considers gender-based violence to fallwithin our Level 3 overall potential risk category. In response it has put in place a range of 59 https://parents.snapchat.com. 177 CONFIDENTIAL mitigation measures. This includes in particular our proactive content moderation which isdesigned to detect and prevent illegal hate speech, including gender-based violence relatedcontent from reaching a broad audience on Snapchat’s in-scope services. Although we do notspecifically document and report on gender-based violence as a category, this type of contentwould primarily fall within our hate speech category, which has an extremely low prevalence(PVP) on Snapchat. We take this matter very seriously, and continue to invest in measures thatprevent this type of content from reaching a broad audience on Snapchat, as well as provide ourusers with tools to report content to Snapchat and law enforcement, and support our communityvia online and in-app support tools, such as Here For You and our Safety Center resources. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate and effective mitigation measures for gender-based violence. There is nochange in this conclusion from our 2023 Report. 4.4.3 Negative Effects on Children We understand that, without mitigations, online platforms could have a negative impact onchildren and Teens. This is a risk we take seriously as Snap’s priority is protecting the safety andwellbeing of our users whilst ensuring they have a positive experience online. Privacy, safety andsecurity are key values of the company and at the core of our value proposition to our users. Likelihood As explained in Snapchat Community as part of our Introduction to this Report, Snapchat is usedby a wide demographic, with 18-24 years making up the highest percentage of users of Snapchat.Nevertheless, there is still a percentage of our users who are Teens (13-17)). Therefore weconsider that children using Snapchat are just as likely to be exposed to the issues identified inthis Report as other members of the Snapchat Community. Several studies have considered the likelihood of underage use of online platforms. For example,in 2022, Ofcom in the UK found that 60% of children aged 8 to 12 use social media with their ownprofiles.60 There were similar reports in Belgium.61 Another example found the percentage ofunderage users on Snapchat to be relatively low (fewer than 4% of 0-11 year olds in the US in2024) compared with other online platforms such as YouTube (28.6%), Netflix (17.2%) and Disney+(15.6%).62 We take a risk based approach to age assurance at present that aligns with industrypractice. We support further proportionate, reasonable and effective industry wide measuressupported by device OS / app store account level age assurance and are proactively workingwith the Commission and others to try to achieve an EU wide approach. We have highlighted our 62 Youth and Social Media, How US Kids and Teens Use Platforms From TikTok to Snapchat to YouTube,EMarketer, March 2024. 61 Réseaux sociaux, règles d'utilisation, intelligence artificielle : comment a évolué l’utilisation des écranschez les jeunes ?, RTBF, May 31, 2024. 60 Children’s Online User Ages - Quantitative Research Study, Ofcom, updated July 2022, url. 178 CONFIDENTIAL measures regarding minors under 13 in the specific mitigation section below (and provided moredetailed information in Section 5.8). All the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). Weconclude that the relative likelihood of a risk of negative effects on children and Teens forSnapchat’s in-scope services compared with other risks varies depending on the underlyingconcern. It is primarily in the Extremely Low Likelihood but rises to Very Low Likelihood fordissemination of harassment and bullying content (given the second highest prevalence ofviolating content in this category relative to other risks) and adult sexual content (given thehighest relative prevalence of violating content in this category relative to other risks) and fornegative effects on data protection rights (as the high volume of personal data we process givesa high risk of some harm if not handled appropriately). Severity As with likelihood above, the severity of harm caused if a particular issue arises depends on theharm caused. However, we take the safety and wellbeing of the youngest members of ourcommunity very seriously and recognise that this group is particularly vulnerable and if aparticular risk materializes, there is an increased risk that the severity of the harm they suffer ishigher. In general, therefore, children and Teens suffer a risk of greater harm from the issues we haveidentified and we have chosen to place the severity of harm arising from an issue that negativelyaffects children in our ‘severe harm’ category. Overall potential risk prioritization Although the relative likelihood for the negative effects on children varies, Snapchat considersthe risk of harm to fall within the severest category. Consequently, Snap considers this to be a Level 1 overall potential risk. There is no change in this assessment from our 2023 Report. Snap's Mitigations Highlights The risk of negative effects on children falls within our highest risk prioritization level. To protectTeens on Snapchat, at a high level, we focus on three core things: 1) mitigating unwanted contact;2) scanning for, detecting and blocking/removing content that violates our Terms (including ourCommunity Guidelines) or the law; and 3) working with law enforcement to help bring criminals tojustice. Snap has dedicated extensive resources to ensuring protections to safeguard the rightsof Teens on the platform, greatly reducing the likelihood of rights infringement. These measuresare set out in Section 5.8 of this Report. 179 CONFIDENTIAL We also work with Trusted Flaggers in the EU, and globally, on child safety issues, as well as our Safety Advisory Board. For more information on this, see Section 6 of this Report. A more detailed run though of our mitigations to protect Teens on Snapchat is set out in Section5.8 (Protection of Minors). Taken together, these mitigations contribute to a safe and responsibleenvironment for young Snapchatters. Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes, Snapchat’s in-scope services have been adapted to include proactive safety by design features and content moderation for Teens. For example: ● We have created a different product experience for Teens and adults. For example, we don’t show sexually suggestive content to Teens. ● All content on Discover has to be appropriate for 13+. ● Regulated goods don’t appear in ads to Teens. ● Snap Map is designed to mitigate particular risks to Teens. For example, location sharing is off by default. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes. For example, our Terms require that all content is appropriate for 13+, we require all users on our platform to be over the age of 13, and we strictly enforce our terms. If we discover that a user is under the age of 13 we will remove their account. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, Yes, specific proactive and reactive moderation procedures to age gate and ensure age-appropriate content (for example restricting Teens access to suggestive content), adjust content settings as designated in Family Center and remove reported content from view. 180 CONFIDENTIAL as well as adapting any relevant decision-making processes and dedicated resources for content moderation. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems take user age into account to provide age appropriate recommendations. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. We have also launched changes to Snapchat’s in-scope services so they no longer display advertisements based on profiling for our under 18 accounts in the EU. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we have specific prevalence testing and transparency reporting for violations, including for example in relation to CSEAI. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with numerous trusted flaggers for child safety who are able to flag other CSEAI or other illegal and violating activities involving Teens. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups e.g. Technology Coalition, WeProtect Global Alliance, EUIF, Alliance and CIPL to better protect minors online. We are actively participating in efforts by the Commission and other stakeholders to develop a EU wide AADC. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. All information on our Privacy and Safety Center or our Support Center is drafted for 13+. For example, Privacy By Product - Privacy Features | Snapchat Privacy provides Teens with ample opportunity to understand the Snapchat features. We also provide Family Center as a resource to Teens and their parents or trusted adults. 181 CONFIDENTIAL Protection of Minors Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. Yes, Snapchat’s in-scope services have been adapted to include proactive safety by design features and content moderation for Teens. We make available robust reporting and enforcement of our terms. Our Family Center - Parental Control For Teens | Snapchat Safety provides Teens and their parents or trusted adults a suite of resources and guidance. Our new parents site provides additional guidance for parents and carers on risks and support.63 Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. Yes, Snap has taken steps to mitigate the risk that (i) its generative AI tools are used for creating illegal or otherwise violating content and (ii) illegal or otherwise violating content created using generative AI tools on any online platform are disseminated on Snapchat’s inscope services. We display an icon in some Lenses that manipulate an image of a Snapchat to make them look younger. Conclusion Although the prevalence of public content that may have negative effects on children on Snap’sin-scope services is generally very low, we recognize that Teens are at risk of greater harm ifexposed and we take the safety and wellbeing of our community, particularly its youngestmembers, very seriously. As such, we have assessed this risk to be in our higher risk prioritizationcategory, Level 1, relative to other risk categories. In response, Snap has put in place a range of mitigation measures. This includes general platformsafeguards such as our Teen friendly terms and support pages, our moderation and enforcementprocesses, our parental tools–Family Center, in-app reporting, and Teen specific contentmoderation and restrictions. Plus, additional safeguards have been put in place to help Teensunderstand and recognize Lenses and ensure that advertisers and advertisements on ourplatform comply with our requirements. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate, and effective mitigation measures for the risk of negative effects on Teens andchildren under 18. In addition, as explained in the Codes section of this Report, we stronglysupport and continue to actively participate in cross-stakeholder efforts to develop an EU 63 https://parents.snapchat.com. 182 CONFIDENTIAL wide AADC and/or guidance on high level of privacy, safety and security under Art 28 of theDSA to assess if further reasonable, proportionate and effective measures are needed foronline platforms, ‘gateways’ (such as device operating systems, app stores and web browsers)and other online services. 4.4.4 Serious Negative Consequences on physical and mental well-being We’ve made it a point to build things differently from the beginning, with a focus on helpingSnapchatters communicate with their close friends in an environment that prioritizes their safetyand privacy. That’s why Snapchat is purposely designed differently from traditional social media.It doesn’t open to a public news feed powered by an algorithm with likes and comments. Instead,as outlined earlier in this report, Snapchat opens to a camera and has five tabs: Camera, Chat,Map, Stories, and Spotlight. Additionally, conversations on Snapchat delete by default to reflectreal-life conversations. Before social media, our fun, spontaneous, and silly interactions withfriends only lived on in our memories. Snapchat is designed to mirror that dynamic, to helppeople feel comfortable expressing themselves without feeling pressure or judgment. We willdiscuss these risks and our mitigations in more detail below. Likelihood All of the risks we track on Snapchat have a low prevalence compared to the prevalence of theseissues elsewhere online and offline. To aid our prioritization, our methodology seeks to assessthe relative likelihood between the risks we track (even though all are low in absolute terms). Snap assesses that serious negative consequences on physical and mental wellbeing are high inlikelihood in the absence of appropriate mitigations. Without mitigations, users of digital platformsmay be exposed to content affecting their mental health, contributing to body dissatisfaction andlow self-esteem. They may also be exposed to content inciting physically harmful activities, suchas dangerous pranks or challenges. On Snapchat, data related to relevant policy enforcements suggests a low prevalence of contentassociated with harm to physical well-being on Snapchat. For example: ● In our 2023 Report, we measured the prevalence (PVP) of self-harm content (including thepromotion or glorification of unhealthy behaviors) and content promoting dangerousactivities to be extremely low. ● We have subsequently observed a significant decrease in the prevalence of all of thesecategories of content. We also monitor community support requests, which are a good indication of the well-being ofour Community. We continue to receive very few user complaints from EU users related todangerous categories that could be considered to have a serious negative effect on physical andmental well-being 183 CONFIDENTIAL We have assessed the amount of average time that minors spend on the features of Snapchatand it is worth noting that the majority of the time spent (i) relates to time spent on surfaces thatare not in scope of the DSA and this Report; and (ii) is focussed on communication with closefriends and family (see the table below which shows that Teens spend the majority of their timeusing Chat and Camera features)). We have also assessed the average time spent, as well as the time of day, that Teens useSnapchat and note that the majority of time is spent in the evening and that average time theaverage times are not excessive. As noted in the mitigations section below, many devices nowinclude well being settings such as turning on bedtime mode by default and providing screentime controls for parents / responsible adults. Over 90% of our community says they feel happy, connected, and comfortable while usingSnapchat.64 Research from University of Chicago’s NORC65 stows that 2 in 3 say messaging withfamily and close friends makes them extremely or very happy. On the other hand, a majority ofteens and young adults feel overwhelmed at the way traditional social media makes them feelpressured to post content that will get lots of likes and comments, or will make them look good toothers. Perhaps most importantly, according to the NORC data, respondents who use Snapchatreport higher satisfaction with the quality of friendships and relationships with family thannon-Snapchatters. As we were finalizing this Report, we were made aware of a new study from the Netherlands(currently in pre-print).66 According to the researchers, it is the first quantitative study thatcompares online platforms on three key factors, impact on well-being, self-esteem and friendshipcloseness. The researchers conducted an extensive 100-day daily diary study among 479adolescents (14-17 years). With respect to Snapchat specifically, the research concludes thatSnapchat is the only platform that positively impacts well-being and Snapchat also has a strongpositive effect on friendships and no net negative effect on self-esteem. It notes in particular: ● Snapchat positively affected friendship closeness and well-being but had no significantimpact on self-esteem. Using WhatsApp had a notably strong effect on friendshipcloseness but no significant effect on well-being and self-esteem. ● The majority of adolescents (60%) experienced unity in negative effects of social media,suggesting that social media use is a contributor to mental health issues. Moreover, 13.6%of adolescents experienced duality in effects, indicating that social media usesimultaneously harms and benefits different dimensions of their mental health. ● The positive and null effects associated with Snapchat and WhatsApp indicate that weshould avoid a blanket condemnation of all social media platforms. 66 Social Media Use Leads to Negative Mental Health Outcomes for Most Adolescents, Amber van der Wal,Ine Beyens, Loes H. C. Janssen, and Patti M. Valkenburg, 2024, url (preprint) 65 https://www.norc.org/about/who-we-are.html. 64 2022 Alter Agents study commissioned by Snap Inc. url. 184 CONFIDENTIAL ● Snapchat scores a 41.4% positive effect on well-being, 23,7% on self-esteem and 71,5% onfriendship closeness, as shown in the figures and table below: 185 CONFIDENTIAL Accordingly, while Snap assesses these risks across digital platforms to be high in the absence ofsafeguards and mitigations, we are encouraged by research and data indicating that ourapproach to mitigating these risks is effective at reducing the likelihood of such negative impactson physical and emotional wellbeing on Snapchat. We continue to assess there to be an Extremely low likelihood of encountering this content on Snapchat. Severity Snap has assessed information published by governments and other third party sources andconsiders that if content that has a negative effect on physical and mental well-being were tomaterialise on an online platform, this could fall within our ‘severe harm’ category. Overall potential risk prioritization Snap's prevalence reports and community feedback suggest that the likelihood of this risk isrelatively low If we would follow our matrix, we would qualify the overall potential of this risk asLevel 3, however, given the importance of this issue, especially in relation to younger users, wehave decided to deviate from the matrix, and marked the potential risk as falling within our Level1 overall potential risk prioritization category. There is no change in this assessment from our2023 Report. 186 CONFIDENTIAL Snap’s Mitigations Highlights Snapchat is intentional about addressing risks to the physical and emotional wellbeing of users.Our Community Guidelines prohibit a range of behaviors and content that may negatively impactwellbeing, including bullying and harassment; content or Lenses that glorify unhealthy behaviorsor promote unrealistic beauty standards; violent or disturbing content, or content that promotesdangerous activities; and content that promotes suicide or self-harm. To combat this, we have put in place a number of specific protections, we will highlight some ofthem here. On Spotlight, we put in place protections for both creators and views: ● Creator protections ○ Users choose whether to post to Spotlight and can choose to disable comments. ○ If comments are not disabled, Spotlight comments are auto-moderated for abusivelanguage before they are viewed by the creator and all comments can bereported to human moderation. This protects the creator from seeing harmfulcomments. ○ Teens are protected on Spotlight by not having their usernames displayed. ○ We limit the recommendation of content from younger users to older users onSpotlight. This is to protect Teens from being contacted by older users. ○ We provide users the ability to post content to Spotlight anonymously. ○ Creators can choose to approve comments on their Spotlight Stories prior topublication. ○ We do not show view-counts on Spotlight below a certain threshold. This is toprevent focus on low view numbers. ○ We aim to distribute content created by Teens to Teens. This is to prevent Teensfrom building a following that is not their own age. ○ Creators are in control of adding hashtags / topics to their videos. This providescreators some control over how their content is categorized. ● Viewer protections ○ Content on Spotlight does not auto-play. ○ We do not have public “favorites”, i.e. a user’s likes and interests are not public. ○ Viewers can “hide” either content or a creator. Subsequently, the user will have alower likelihood of seeing content of such nature or content from the creator thathas been “hidden”. ○ We survey a subset of our users quarterly to understand whether they find theirtime spent on our experience entertainment and satisfactory. We use this to trackwhether our product changes are improving viewers' overall perception of theapp. 187 CONFIDENTIAL ○ We provide a diversity of perspectives. We have multiple programs to foster amore diverse content community and surface different perspectives (e.g. BlackCreator Accelerator program). ○ We ensure there is always a large mix of content from creators from viewers’home country and content in the language in which they have set their device. ○ We add diversity to every viewer’s feed in terms of the account they see, and thecategories of content we surface to them. This prevents users from entering anecho chamber or filter bubble of seeing the same content repeatedly. We usemachine learning to understand content categories and diversify it. We believe the above measures contribute to the well-being of both creators and viewers andcreates a more pleasant experience. Similar measures are in place for Public Profiles (which arecurrently only available for adult accounts). For example, we allow users with access to a PublicStory to turn off all Story Reply messages so they don’t see messages from users who reply totheir Stories. We also give users control over Story Replies and filter out words they don’t want tosee. Users can input words that they don’t want to receive in the Story Replies from theirsubscribers. If a Story Reply contains an inputted word, the user does not receive the story reply(and any other story replies) from the sender. Additionally, we allow creators to block repliers orreport them. In addition, as we have explained in the mitigations section of this Report, in particular Section5.8 (Protection of Minors), teenagers, parents and other responsible adults are able to set timelimits for their teenagers, amongst other controls, via the device operating system’s family tools (e.g. Google Family Link, Apple device parental controls and Family Sharing controls and Microsoft Family Safety). Mobile devices now also commonly provide default settings for late night usage, such as bedtime modes that turn off device and app notifications and turn thescreen the black and white to encourage sleep. We have also undertaken considerable efforts to stay apprised of users' wellbeing: ● On Safer Internet Day, 6 February 2023, we launched our inaugural Digital Well-Being Index (DWBI), a measure of Generation Z’s online psychological well-being. To gain insight into how teens and young adults are faring online – across all platforms and devices, notjust Snapchat – and to help inform our Family Center and the broader online ecosystem,we polled more than 9,000 people across three age demographics in six countries. ● Not surprisingly, the research showed that social media plays a major role in Gen Z’sdigital well-being, with more than three-quarters (78%) of respondents saying social mediahad a positive influence on the quality of their lives. More information about that research can be found here. ● We have repeated this research in 2024 and will be publishing the results in September.Further information can be found in Section 6.6 (DWBI Initiative). 188 CONFIDENTIAL Specific Mitigations In the table below we indicate the specific measures we have taken to mitigate this risk in respectof Snapchat’s in-scope services, using the defined list of mitigations set out in Article 35 of theDSA. The primary purpose of the below table is to indicate whether each specific mitigationcategory applies to this risk and the descriptions are illustrative rather than exhaustive. As manyof our mitigations apply to all of the risks assessed in this Report, to reduce duplication in thisReport, each row in the tables provides a link in the left hand column to a full summary of thespecific mitigation in Section 5 of this Report which explains in more detail how each mitigationoperates to reduce the risk. DSA Mitigation Applies to this risk? Snapchat Design and Function Adapting the design, features or functioning of their services, including their online interfaces. Yes. Snapchat has incorporated a host of design, features and functions to address this risk. Starting with the aforementioned decision to open to the camera and not a news feed. This encourages self expression, communication, and exploration through our AR Lenses. Snapchat and third parties have created Lenses centered on movement, fitness, yoga poses, breathing activities. In addition to this, there are several partnered lenses that prompt Snapchatters to talk about wellness, mental health and their experiences. Our user generated content feature, Spotlight, has both creator and viewer protections in place. Terms and Enforcement Adapting their terms and conditions and their enforcement. Yes. Our Community Guidelines prohibit a range of behaviors and content that may negatively impact wellbeing, including bullying and harassment, content or Lenses that glorify unhealthy behaviors or promote unrealistic beauty standards, violent or disturbing content, or content that promotes dangerous activities, and content that promotes suicide or self-harm. Our Community Guidelines are enforced. Moderation Adapting content moderation processes, including the speed and quality of processing notices related to specific types of illegal content and, where appropriate, the expeditious removal of, or the disabling of access to, the content notified, in particular in respect of illegal hate speech or cyber violence, as well as adapting any relevant Yes, specific proactive and reactive moderation procedures to prevent bullying and harassment, content or Lenses that glorify unhealthy behaviors or promote unrealistic beauty standards, violent or disturbing content, or content that promotes dangerous activities, and content that promotes suicide or self-harm. 189 CONFIDENTIAL decision-making processes and dedicated resources for content moderation. Algorithmic Systems Testing and adapting their algorithmic systems, including their recommender systems. Yes, our algorithmic systems do not categorize or recommend content that our Community Guidelines prohibit. Our Content Guidelines for Recommendation Eligibility - Snap Inc. further describe how sensitive and disturbing content is demoted for distribution on Spotlight and Discover. For example, glorification of violence is not suggested content to users on Spotlight or Discover and any discussion on self-harm, including eating disorders is demoted to users based on their age, location, or personal preferences. Advertising Systems Adapting their advertising systems and adopting targeted measures aimed at limiting or adjusting the presentation of advertisements in association with the service they provide. Yes, other mitigations listed here also apply to our Advertising Systems. For example, ads for diet and fitness products or services must not demean the user, or shame anyone on the basis of body shape or side. Risk Detection and Management Reinforcing the internal processes, resources, testing, documentation, or supervision of any of their activities in particular as regards detection of systemic risk. Yes, we have specific prevalence testing and transparency reporting for harassment and bullying and self-harm and suicide and other prohibited content on Snapchat that may impact users mental wellbeing. Trusted Flaggers Initiating or adjusting cooperation with trusted flaggers in accordance with Article 22 and the implementation of the decisions of out-of-court dispute settlement bodies pursuant to Article 21. Yes, we cooperate with trusted flaggers in relation to illegal hate speech and child safety. Codes and Crisis Protocols Initiating or adjusting cooperation with other providers of online platforms or of online search engines through the codes of conduct and the crisis protocols referred to in Articles 45 and 48 respectively. Yes, we cooperate with other providers through various industry groups e.g. EUIF. In 2017 Snap joined FSM and has signed the FSM Code of Conduct which aims to protect users from content offered on digital services that could endanger or impair their development. Transparency Taking awareness-raising measures and adapting their online interface in order to give recipients of the service more information. Yes, we provide guidance on our terms, harms, moderation and enforcement practices (see the Annex), as well as how to and how to get help in our Safety Center. Protection of Minors Yes, we have protective measures to ensure age appropriate content and our Family Center offers 190 CONFIDENTIAL Taking targeted measures to protect the rights of the child, including age verification and parental control tools, tools aimed at helping minors signal abuse or obtain support, as appropriate. resources and guidance. Our new parents site provides additional guidance for parents and carers on risks and support.67 Device operating systems commonly provide settings for Teens and their parents/responsible adults to manage screentime and late night use. Content Authenticity Ensuring that an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces, and, in addition, providing an easy to use functionality which enables recipients of the service to indicate such information. General content authenticity measures. We are displaying an icon in some Lenses that manipulate an image of a Snapchat to make them look younger. Conclusion Given the heightened potential for negative consequences on physical and mental well-beinginherent to online platforms, specifically social media, despite the prevalence on Snapchat beinglow, we consider the overall potential risk prioritization to be Level 1. In response, Snap has made deliberate design and policy decisions to reduce the potential forharm on Snapchat. Snap has implemented numerous protections for both creators and viewers ofSpotlight content and undertaken considerable efforts to understand users’ wellbeing onSnapchat and other platforms. We have concluded therefore that Snapchat’s in-scope services have reasonable,proportionate, and effective mitigation measures for the risk of negative effects on physicaland mental well-being. There is no change in this conclusion from our 2023 Report. 67 https://parents.snapchat.com. 191 CONFIDENTIAL 5\. Specific Mitigations Article 42(4)(b) of the Digital Services Act requires providers of Very Large Online Platforms toreport on the specific mitigation measures that they have put in place pursuant to Article 35(1) ofthe DSA. Article 35(1) of the Digital Services Act requires providers of Very Large Online Platformsto put in place reasonable, proportionate and effective mitigation measures, tailored to thespecific systemic risks identified pursuant to Article 34 of the DSA, with particular consideration tothe impacts of such measures on fundamental rights, including where applicable the definedcategories of measures set out in Article 35(1)(a)-(k). In Section 4 of this Report above, we reported on our: (i) assessment of the specific systemic risksapplicable to Snapchat’s in-scope services; (ii) summary of the mitigation measures that Snap hasin place tailored to those risks and (iii) conclusion as to whether those mitigation measures arereasonable, proportionate and effective. In this Section 5, we have provided details of the specificmitigations that Snap has put in place, as are summarized in Section 4, to comply with ourobligation under Article 42(4)(b). This section 5 uses the defined categories of measures set out in Article 35(1) to detail thesemeasures, with the following subsections: ● Snapchat Design and Function (Article 35(1)(a)) ● Terms (Article 35(1)(b)) ● Transparency (Article 35(1)(i)) ● Moderation (Article 35(1)(c)) ● Enforcement (Article 35(1)(b)) ● Algorithmic Systems (Article 35(1)(d) ● Advertising Systems (Article 35(1)(e) ● Protection of Minors (Article 35(1)( j) ● Content Authenticity (Article 35(1)(k) ● Trusted Flaggers (Article 35(1)(g)) ● Dispute Settlement Bodies (Article 35(1)(g)) ● Codes and Crisis Protocols (Article 35(1)(h)) Note that our measures to reinforce the internal processes, resources, testing, documentationand supervision of our activities as regards to the ongoing detection and management of DSArisks, as referred to in Article 35(1)(f), is set out in Section 6 of this Report. 192 CONFIDENTIAL 5.1 Snapchat Design and Function 5.1.1 Adaptations and Mitigations As a result of our privacy and safety by design approach (described in Section 6.3 of this Report),Snapchat was designed from the outset with core features and functionalities that mitigate therisks described in Snap’s Risk Assessment Report. As we have made these key foundational design decisions from day one. We hear fromSnapchatters about the benefits of these choices all the time, as well as consulting with expertand teen stakeholders (such as those forming part of our Safety Advisory Board and TeenCouncil) and we believe that these foundational design decisions directly influence those results. Although not all of the features listed below are in scope of the risk assessment, we haveincorporated a summary of holistic mitigations that we have put in place to demonstrate ourprivacy and safety by design approach below: Friends First, by default users need to accept bi-directional friend requests or already have each other intheir contact book to start communicating directly with each other. This design decision addsfriction and prevents users from communicating with each other prior to accepting a friendrequest or being in one’s contact book. Private friend lists Second, once users have accepted friend requests, the friend lists remain private. Snapchat doesnot disclose the friend lists of users to other users, nor do we expose the total number of friendsthat a user has. This protects the privacy of the user and their friends. On most other platformsfriend lists are public by default or there is an option to share them publicly. These types offeatures create the ability for strangers to contact vulnerable groups (e.g. younger users). Open to the Camera not a feed Third, Snapchat opens to the Camera and invites people to express themselves. At the surface,this may sound like a small design decision, but it directly impacts the user behavior on theplatform. Instead of inviting users to scroll a feed of content, the invitation to users is to expressthemselves, live in the moment and share a moment with their close friends. Stories are by default set to be viewable by friends, not the public Fourth, once users decide to share a Snap via My Story, by default only friends can view it.Snapchatters can choose to share to everyone, only to friends, or to a customized few. Thisemphasis on sharing with friends and giving users controls over who can view their content are inline with how Snap takes into account privacy and safety when designing its features. 193 CONFIDENTIAL No focus on public vanity metrics Fifth, once a user posts to their Story, we don’t show vanity metrics, such as likes on that Storycontent. The goal is not to create a popularity contest around who has the most friends or likes.68 The design choice is to provide all users with a more authentic form to express themselves. As a result of our privacy and safety by design approach, each of Snapchat’s in-scope serviceshas been designed with features and functionalities that mitigate the risks described in the RiskAssessment Results section above. Spotlight and Discover Simplified Snapchat As noted in the What’s New and Introduction sections of this Report, we are consideringsimplifying the Snapchat application from 5 to 3 tabs. Rather than having the current separateSpotlight and Discover tabs, this envisages a single unified tab for more public content to theright of the Camera. We plan to run some tests later this year to assess how it performs. This isprimarily a cosmetic change, and it should not impact the mitigations including with respect to thedesign and function of what is currently Spotlight and Discover. For now we have continued todescribe these mitigations separately for Spotlight and Discover below. Spotlight Spotlight offers creators at all stages of their career a variety of opportunities and tools to helpthem grow their audiences, build sustainable businesses and make content creation a full-timecareer. The content shown in Spotlight is personalized to provide viewers with a more relevantexperience, that ‘spotlights’ the best content on Snapchat. We have made following designdecisions to protect our creators and users: ● Creator protections ○ Users can post to Spotlight and choose to disable comments. ○ If comments are not disabled, Spotlight comments are auto-moderated for abusivelanguage before viewed by the creator and all comments can be reported tohuman moderation. This protects the creator from seeing harmful comments. ○ Adults cannot comment on Teen’s Stories on Snapchat. ○ Teens are protected on Spotlight by not having their usernames displayed. ○ We restrict Teens' ability to reach a large audience on Spotlight to prevent olderusers from seeing content from younger users. This is to protect Teens from beingcontacted by older users. ○ Creators can choose to approve comments on their Spotlight Stories prior topublication 68 See also our More Snapchat campaign https://www.moresnapchat.com. 194 CONFIDENTIAL ○ We do not show views on Spotlight below a certain number of views. This is toprevent pressure over low view numbers. ○ We aim to distribute content created by minors to minors. This is to prevent minorsfrom building a following that is not their own age. ○ Creators are in control of adding hashtags / topics to their videos. This givescreators control over how their content is categorized. ● Viewer protections ○ Content on Spotlight does not auto-play ○ We do not have public “favorites”, i.e. a user’s likes and interests are not public ○ Viewers can “hide” either content or a creator. Subsequently, the user will not seemore content of such nature or content from the creator that has been “hidden”. Discover Discover is dedicated to Creator Stories, which includes Media Partner content and some usergenerated content from popular users (“Creator Content”). The Creator Content that appears onDiscover includes Public Stories from Snap Stars and other users who meet a follower countthreshold. Similar to Spotlight, we made following design decisions to protect our creators andusers: ● Creator protections ○ Viewer comments are not typically available on Discover content. Wherecomments are enabled, they are subject to auto-moderation. Creators and otherusers can report comments, which leads to human review. They can also blockcommenters which will prevent them from ever seeing the blocked users’comments again on any content. ○ We do not show “views” on Discover Stories. This protects creators from feelingembarrassed or being subject to ridicule due to low number of views. ○ Content published by creators has a limited publication duration (which may bechanged by creators with a Snapchat+ subscription). This protects creators byensuring their content is not available forever. ○ Creators are free to re-publish new and saved stories at any time, provided it doesnot violate the law or our Terms. ● Viewer protections ○ Content on Discover does not auto-play ○ We do not have public “favorites”, i.e. a user’s likes and interests are not public Public Profile Users with a Public Profile can post Public Stories that are publicly viewable for all Snapchatters.Additionally, Snapchatters can permanently showcase their Public Stories and Spotlights on their profile. Snapchatters can Follow a Public Profile from Spotlight, Snap Map, Discover, or by using the Follow button on a Public Profile. Unlike friend requests to non-Public Profile owners, Public 195 CONFIDENTIAL Profile owners will not receive a notification for new followers. We have made the followingdesign decisions to protect users with Public Profiles: ● Users can easily delete all of their public content. We allow users to delete all of theirpublic content in a single tap. We delete any and all content they added to their Public Profile and that is publicly viewable. Our public options are in fact options. If Snapchatters are not or no longer interested in being a creator and showcasing content publicly, theycan simply choose to not add to their Public Profile, post to their Public Story, or share toSpotlight and to the Snap Map. ● We give users control over content that is publicly viewable by allowing users to hide orshow their Spotlights on their Public Profile both at the time of submission and aftersubmission. ● Public Profile users can turn off remixes. We allow users to decide whether their publiccontent can be remixed by other users. ● We educate users on their public options and attribution controls. When users first tap ontheir public profile, public story, and spotlight/snap map posting, we show themeducational modals that educate them about the public option. ● To ensure that users are aware when they become friends with another user so that theycan control what data that user has access to, we send notifications to the user when theybecome friends with another user (bi-directional add has occurred). ● Only users with a declared age of 18 or older can have a Public Profile and Public Story.Viewers cannot distinguish between users without Public Profiles (under 18) and userswith Public Profiles (18+) who have not edited the Profile in any way. Note: Since completion of the 2024 Risk Assessment we announced that we are reviewing a 16-17 Public Profile experience with specific mitigations for this age group. This product is not yet available in the EU. It will not be rolled out to the EU until we have finalized our review and completed an update to our risk and mitigation assessments as needed. ● We give users control over their ability to be contacted. We allow users with access to aPublic Story to turn off all Story Reply messages so they don’t see messages from userswho reply to their Stories. We also give users control over Story Replies and filter outwords they don’t want to see. Users can input words that they don’t want to receive in thestory replies from their fans. If a story reply contains an inputted word, the user does notreceive the story reply (and any other story replies) from the sender. Additionally, we allowcreators to block repliers or report them. We have also built in protections for users who engage with a Public Profile. For example, weinform users before they send a Story Reply to a creator that the creator could quote the replyand make it publicly viewable (with the replier’s first name and Bitmoji). We also limit unwarrantedconnections between younger users and bad actors on the platform. 196 CONFIDENTIAL Snap Map We filter Stories posted from users with new accounts so they do not feature on Snap Map.Further, content posted to Public Stories will only show on Snap Map with a clear location if thereare multiple users posting in a short time nearby and a percentage of those posters are non-Teenaccounts. Teens in Europe also don’t have the ability to post to Public Stories which means thattheir Snaps are not eligible for the Snap Map. When younger individuals (under 16 in EEA) useSnap Map, Snap collects and uses precise location data only for the purpose of providing thefeature to the Snapchatter and only stored for a short period. These controls prevent illegitimate use of Snap Map and protects Teens from exploitation. Snap Map has numerous additional design choices in place to make Snap Map a safer space forour community. These include: ● Opt-in and only Friend sharing. Given the sensitivity of geolocation data, users mustgrant Snap location permission via a just-in-time choice option, and even after grantingthat permission users must additionally opt-in to sharing their location with others on theSnap Map. Users must affirmatively opt-in to share their location with friends. Locationsharing is disabled by default, and sharing preferences with friends can be easilychanged by users at any time in app settings. ● No option to share location with strangers. We want location sharing on Snap Map to belimited to engagement with friends on Snapchat. We also want to ensure user safety bynot broadcasting a user’s location to others who are not friends of the user. Therefore,users cannot share their location with strangers. In Settings, users can choose to sharetheir location with their friends, or a subset of friends only. There is no option to sharetheir location with non-friends. Friendship must be bi-directional. ● Permission and prompts. By default, users are not sharing their location with any friends,as all users are defaulted to “Ghost Mode”. This was to ensure that location sharing wouldbe understood by users before activation, in particular younger users so they could makeinformed choices about whether to use Snap Map, whether to share their location and, ifso, with whom to share it. Snapchatters can update: (1) whether they are sharingbackground location or only while using location; and (2) who among their friends can seetheir location at any time right from the settings gear in the Map. ● All Friends Confirmation. We check if users really want to share with all Friends the firsttime they choose the “My Friends” option with an in-app pop-up dialogue. This extrareminder appears once per account when the user selects or switches to All Friendssharing (if user selects “not now”, the reminder will still be shown the next time the userattempts to switch to All Friends): 197 CONFIDENTIAL ● Location Sharing Change Confirmations. When users change their location accesssettings (e.g. from “While Using the App” to “Always”, we display in-app alerts to confirmthe change (e.g. "Your location now updates in real-time"). These are shown on the firsttime you come back to the app every time you change your access settings. ● Ghost Mode Exit Confirmation. Similar to the location change confirmations above, wealso display a "You're on the Map!" reminder when the user exits Ghost Mode and is visible toany Friends they are sharing location with. ● Device OS icon and reminders. We want users to know at the moment wheneverSnapchat is accessing their device location data. Device OS automatically displays arecognizable icon for users to know whenever an app is accessing device location data.The icon is consistent across all apps on the device that accesses location data, so itshould be familiar and instantly understood by users across age groups. ● Location Sharing Reminders: We also want users, especially those who don’t regularlyengage with the Map/Map settings, to be regularly reminded their location is visible andto which friends. ● Auto-expiration of Last Active Location. Where users have selected to share locationonly while using the app, we auto-remove users' location from the Map if they have notopened the app after 24 hours. ● Creator protections ○ Currently there is no comments on Snap Map content ○ We do not show “views” for Stories on Snap Map. Protects creators from feelingembarrassed or being subject to ridicule due to low number of views. ○ Content published by creators has a limited publication duration (which may bechanged by creators with a Snapchat+ subscription. This protects creators byensuring their content is not available forever). ○ Creators are free to re-publish new and saved stories at any time, provided it doesnot violate the law or our Terms. ● Viewer protections ○ We do not have public “favorites”, i.e. a user’s likes and interests are not public . Lenses Lenses (in popular language often dubbed as ‘filters’) are created by a relatively limited numberof community developers, and Snap’s internal Lens Team. Our Lenses are designed withprivacy-and-safety-by design principles in mind. For example, Lenses require object detectionrather than facial identification. Lenses can tell what is or isn’t a face, they do not identify specific 198 CONFIDENTIAL faces, limiting data processing for the use of Lenses. Snap does also not use any data collectedby Lenses to customize the content that the user sees in Spotlight or Discover, nor is any datacollected for advertising purposes. Besides, voice data collection of Snapchatters in the EU is offby default; it is only used to provide the service. Snap also designs every Lens with race, gender, ethnicity and cultural norms in mind. Snapleverages its ever-growing diversity training datasets, as well as feedback from communitymembers. If a Lens does not resonate with our community, as expressed through a high ratio ofuser reports, we take that feedback into consideration and will re-review the Lens with a goal toleave as-is, modify, or remove. Advertising We have also put in place risk mitigation measures for our advertising efforts. We preventadvertisers from manipulating small audiences with microtargeted campaigns, particularly forpolitical ads. We do so by requiring a specific minimum audience of Snapchatters to be targeted (including Dynamic Ads on Snapchat | Snapchat for Business). This prevents microtargeting that can influence voters politically or push targeted misinformation to certain populations. Ouradvertising systems also do not use ‘special category’ personal data to target ads and we requireadvertisers to provide additional information for political ads. 5.1.2 Integrations with other mitigations On Snapchat we have also adapted our features to integrate with our other risk mitigationsdescribed in this Specific Mitigations section of the Report, for example: Terms All public content must adhere to our Terms, for example the content must be suitable for 13+, inorder to be featured or receive broad distribution on Spotlight and Discover. This is explained inSections 5.2 (Terms) and 5.3 (Transparency). Content Moderation We moderate content on Snapchat in a number of ways to mitigate the risks of users beingexposed to harmful and illegal content. This is explained in Sections 5.4 (Content Moderation)and 5.5 (Enforcement). Content Distribution We have put in place risk mitigation measures to restrict the distribution of harmful content onSnap. For example: ● Content that is Sexually Suggestive and Sensitive (e.g., potentially-disturbing human bodyimagery, violence, horror, etc.) is not distributed to Teens. ● Spotlight Comments with abusive language are removed. ● Ranking avoids ‘filter bubbles’ through demotion, ensuring similar content isn’t 199 CONFIDENTIAL sequentially recommended to Snapchatters in Discover or Spotlight. This is explained in Sections 5.6 (Algorithmic Systems) and 5.7 (Advertising Systems). 5.1.3 Online Interface Design Process Snap implemented a process and governance around online interface design. 5.1.4 Online Design Principles Snap also established Online Design Principles which prohibit the use of dark patterns \&misleading nudge techniques. 5.1.5 Conclusion From day one, Snap has made conscious design decisions to mitigate systemic risks fromoccurring on its platform, including privacy and safety by design decisions such as shorterretention periods, default Story visibility to just friends, not promoting likes on a user’s Story, nothaving public friend lists, and maintaining proactive content policies. Snap has implementedadditional mitigation measures as further outlined in the remainder of this Report. As explained in Section 4, we have concluded that the adaptations made by Snap to thedesign, features and functioning of Snapchat’s in-scope services, in combination with theother mitigations explained in this Section 5, are reasonable, proportionate and effectivemitigation measures for the risks identified. 5.2 Terms 5.2.1 Introduction This document outlines Snap’s protocol for communicating its platform Terms and Conditions tousers, in compliance with the requirements of Regulation (EU) 2022/2065 on a Single Market ForDigital Services and amending Directive 2000/31/EC (the “Digital Services Act” or “DSA”), inparticular with regard to Articles 14 and 27. Snap publishes Terms and Conditions with concise summaries in clear, easily understandable,unambiguous language, in a publicly available, easily accessible, and machine-readable format.Snap’s Terms and Conditions include detail on: how use of the service may be restricted; contentmoderation policies and procedures; information on the use of algorithms, the parameters andcriteria behind recommender system functioning, and how to adjust them; instances when useraccess and/or content may be restricted, suspended or terminated; and instructions on theinternal complaint handling system. This information is primarily provided in Snap’s Terms ofService, which are translated into all official EU member state languages. 200 CONFIDENTIAL In addition to the Terms of Service, Snap also publishes Community Guidelines, Privacy Policies,Product Specific Terms and Advertising Policies. Snap's Community Guidelines elaborate onrestrictions of the use of Snapchat related to: Sexual Content, Harassment \& Bullying, Threats,Violence \& Harm, Harmful False or Deceptive Information, Illegal or Regulated Activities, andHateful Content, Terrorism, and Violent Extremism. Snap’s Terms of Service prohibit any use ofSnapchat to conduct illegal activities. 5.2.2 Terms and Conditions Snap publishes a series of policies that make up our Terms and Conditions: ● Terms of Service ● Community Guidelines ● Privacy Policy ● Advertising Policies Each of our Terms and how they mitigate each of the DSA risk categories is explained below. Terms of Service Snap publishes its Terms of Service on the website. It is easily accessible via search engine andmachine readable. The Terms of Services include information on: ● Restrictions imposed on use of services (these are elaborated upon in Snap’s CommunityGuidelines, Privacy Policy, and Advertising Policy) ● Policies, procedures, measures and tools used for the purpose of content moderation,including algorithmic decision-making and human review ● Rules of procedure of Snap’s internal complaint handling system and available remediesand redress mechanisms. ● Main parameters used in Snap’s recommender systems. 201 CONFIDENTIAL Restrictions Imposed The following restrictions are included in Snap’s Terms of Service (as at August 2024): Who Can Use the Services Our Services are not directed to children under the age of 13, and you must confirm that you are13 years or older to create an account and use the Services Respecting the Services and Snap’s Rights You must also respect Snap’s rights and adhere to the Snapchat Brand Guidelines, Bitmoji BrandGuidelines, and any other guidelines, support pages, or FAQs published by Snap or our affiliates. Respecting Others' Rights You therefore may not use the Services, or enable anyone else to use the Services, in a mannerthat violates or infringes someone else’s rights of publicity, privacy, copyright, trademark, or otherintellectual property right. Safety By using the Services, you agree that you will at all times comply with these Terms, including ourCommunity Guidelines and any other policies Snap makes available in order to maintain thesafety of the Services. Content Moderation The following Content Moderation information is included in Snap’s Terms of Service. Much of the content on our Services is produced by users, publishers, and other third parties.Whether that content is posted publicly or sent privately, the content is the sole responsibility ofthe user or entity that submitted it. Although Snap reserves the right to review, moderate, orremove all content that appears on the Services, we do not review all of it. So we cannot — and 202 CONFIDENTIAL do not — guarantee that other users or the content they provide through the Services will complywith our Terms, Community Guidelines or our other terms, policies or guidelines. You can read more about Snap’s approach to content moderation on our Support Site. Users can report content produced by others or others’ accounts for violation of our Terms,Community Guidelines or other guidelines and policies. More information about how to report content and accounts is available on our Support Site. We may restrict, terminate, or temporarily suspend your access to the Services if you fail tocomply with these Terms, our Community Guidelines or the law, for reasons outside of ourcontrol, or for any other reason. That means that we may terminate these Terms, stop providingyou with all or any part of the Services, or impose new or additional limits on your ability to useour Services. For example, we may deactivate your account due to prolonged inactivity, and wemay reclaim your username at any time for any reason. And while we’ll try to give youreasonable notice beforehand, we can’t guarantee that notice will be possible in allcircumstances. Before we restrict, terminate or suspend your access to the Services, we will take into account allrelevant facts and circumstances apparent from the information available to us, depending onthe underlying reason for taking that action. For example, if you violate our CommunityGuidelines we consider the severity, frequency, and impact of the violations as well as theintention behind the violation. This will inform our decision whether to restrict, terminate orsuspend your access to the Services and, in the event of suspension, how long we suspend youraccess. You can find out more about how we assess and take action against misuse of our Services on our Support Site. Internal Complaint Handling The following information on Snap’s Internal Complaint Handling is included in Snap’s Terms ofService. Where we restrict, terminate or suspend your access to the Services for violation of ourCommunity Guidelines, we will notify you and provide an opportunity for you to appeal. We hope you’ll understand any decisions we make about content or user accounts, but if youhave any complaints or concerns, you can use the submission form available here or useavailable in-app options. If you use this process, your complaint must be submitted within sixmonths of the relevant decision. Upon receiving a complaint, we will: ● ensure the complaint is reviewed in a timely, non-discriminatory, diligent andnon-arbitrary manner; ● reverse our decision if we determine our initial assessment was incorrect; and 203 CONFIDENTIAL ● inform you of our decision and of any possibilities for redress promptly. Recommender Systems The following information on Snap’s Recommender Systems is included in Snap’s Terms ofService. Our Services provide a personalized experience to make them more relevant and engaging foryou. We will recommend content, advertising and other information to you based on what weknow and infer about your and others' interests from use of our Services. It is necessary for us to handle your personal information for this purpose, as we explain in our Privacy Policy. You can find more information on personalized recommendations on our Support Site. Community Guidelines In our Community Guidelines, which are explicitly incorporated into our Terms of Service, weprovide further guidance on the categories of illegal content, and content that Snap deems inviolation of its Terms. The Community Guidelines are easily accessible via Search Engine and inSnap’s Transparency Center and are machine readable and easily understandable. Our Community Guidelines are broken up into the following sections: Sexual Content,Harassment \& Bullying, Threats, Violence \& Harm, Harmful False or Deceptive Information, Illegalor Regulated Activities, and Hateful Content, Terrorism, and Violent Extremism. These categories have been fine tuned over many years of content moderation on Snapchat, andencompass the illegal content that we have encountered on Snapchat over the years. For ease ofreference we have incorporated a more detailed breakdown of each category (as at August2024) below. Sexual Content ● We prohibit any activity that involves sexual exploitation or abuse of a Teen,including sharing child sexual exploitation or abuse imagery, grooming, or sexualextortion (sextortion), or the sexualization of children. We report all identifiedinstances of child sexual exploitation to authorities, including attempts to engage insuch conduct. Never post, save, send, forward, distribute, or ask for nude or sexuallyexplicit content involving anyone under the age of 18 (this includes sending orsaving such images of yourself). ● We prohibit promoting, distributing, or sharing pornographic content, as well ascommercial activities that relate to pornography or sexual interactions (whetheronline or offline). ● Breastfeeding and other depictions of nudity in non-sexual contexts are generallypermitted. 204 CONFIDENTIAL ● Additional guidance on sexual conduct and content that violates our Community Guidelines is available here. These Terms make clear to Snapchatters the extent to which sexual content is prohibited. Thisreduces the likelihood of several risks falling with categories identified by the DSA, including inparticular: (i) the dissemination of child sexual abuse material and adult sexual content inCategory 1, (ii) the right to human dignity and child rights in Category 2 and (iii) negative effectson public health, Teens and gender-violence in Category 4. Harassment and Bullying ● We prohibit bullying or harassment of any kind. This extends to all forms of sexualharassment, including sending unwanted sexually explicit, suggestive, or nudeimages to other users. If someone blocks you, you may not contact them fromanother Snapchat account. ● Sharing images of a person in a private space — like a bathroom, bedroom, lockerroom, or medical facility — without their knowledge and consent is prohibited, as issharing another person’s private information without their knowledge and consentor for the purpose of harassment (i.e., “doxxing”). ● If someone is depicted in your Snap and asks you to remove it, please do! Respectthe privacy rights of others. ● Please also do not harass another Snapchatter by abusing our reportingmechanisms, such as intentionally reporting content that is permissible. ● Additional guidance on how bullying and harassment violate our Community Guidelines is available here. These Terms make clear to Snapchatters that the extent to which harassment and bullying isprohibited. This reduces the likelihood of several risks falling with categories identified bythe DSA, including in particular: (i) the right to human dignity, private life and data protection,and child rights, in Category 2 and (ii) negative effects on public health, Teens, andgender-violence, as well as serious negative consequences to a person’s physical andmental well being in Category 4. Threats, Violence and Harm ● Encouraging or engaging in violent or dangerous behavior is prohibited. Neverintimidate or threaten to harm a person, a group of people, or someone’s property. ● Snaps of gratuitous or graphic violence, including animal abuse, are not allowed. ● We don’t allow the glorification of self-harm, including the promotion of self-injury,suicide, or eating disorders. ● Additional guidance on threats, violence, and harm that violate our Community Guidelines is available here. 205 CONFIDENTIAL These Terms make clear to Snapchatters the extent to which threats and violence areprohibited. This reduces the likelihood of several risks falling with categories identified by theDSA, including in particular: (i) the right to human dignity and property, and child rights, inCategory 2, (ii) negative effects on civic discourse and public security in Category 3 and (iii)negative effects on public health, Teens, and gender-violence, as well as serious negativeconsequences to a person’s physical and mental well being in Category 4. Harmful, False, or Deceptive Information ● We prohibit spreading false information that causes harm or is malicious, such asdenying the existence of tragic events, unsubstantiated medical claims, underminingthe integrity of civic processes, or manipulating content for false or misleadingpurposes (whether through generative AI or through deceptive editing). ● We prohibit pretending to be someone (or something) that you’re not, or attemptingto deceive people about who you are. This includes impersonating your friends,celebrities, public figures, brands, or other people or organizations for harmful,non-satirical purposes. ● We prohibit spam, including pay-for-follower promotions or other follower-growthschemes, the promotion of spam applications, or the promotion of multilevelmarketing or pyramid schemes. ● We prohibit fraud and other deceptive practices, including the promotion offraudulent goods or services or get-rich-quick schemes, or imitating Snapchat orSnap Inc. ● Additional guidance on harmful false or deceptive content that violates our Community Guidelines is available here. These Terms make clear the extent to which harmful false or deceptive information isprohibited. This reduces the likelihood of several risks falling with categories identified by theDSA, including in particular: (i) the dissemination of harmful false information, fraud and spam inCategory 1, (ii) the right to human dignity, private life and data protection, and child rights inCategory 2, (iii) negative effects on democratic and electoral processes, civic discourse andpublic security in Category 3 and (iv) negative effects on public health, Teens, andgender-violence, as well as serious negative consequences to a person’s physical and mentalwell being in Category 4. Illegal or Regulated Activities ● Don’t use Snapchat to send or post content that’s illegal in your jurisdiction, or forany illegal activity. This includes promoting, facilitating, or participating in criminalactivity, such as buying, selling, exchanging, or facilitating sales of illegal orregulated drugs, contraband (such as child sexual exploitation or abuse imagery),weapons, or counterfeit goods or documents. It also includes promoting orfacilitating any form of exploitation, including sex trafficking, labor trafficking, orother human trafficking. 206 CONFIDENTIAL ● We prohibit the illegal promotion of regulated goods or industries, includingunauthorized promotion of gambling, tobacco or vape products, and alcohol. ● Additional guidance on prohibited illegal or regulated activities that violate our Community Guidelines is available here. These Terms make clear the extent to which illegal or regulated activities are prohibited. Thisreduces the likelihood of several risks falling with categories identified by the DSA, including inparticular: (i) the dissemination of illegal content, including child sexual abuse material andother types of misuse of Snapchat for criminal offences, and the conduct of illegal activities,such as the sale of products or services prohibited by European Union or Member State law,including dangerous or counterfeit products, or illegally-traded animals in Category 1, (ii) theright to property and child rights in Category 2, (iii) negative effects on public security inCategory 3 and (iv) negative effects on public health and Teens, as well as serious negativeconsequences to a person’s physical and mental well being in Category 4. Hateful Content, Terrorism, or Violent Extremism ● Terrorist organizations, violent extremists, and hate groups are prohibited from usingour platform. We have no tolerance for content that advocates or advancesterrorism or violent extremism. ● Hate speech or content that demeans, defames, or promotes discrimination orviolence on the basis of race, color, caste, ethnicity, national origin, religion, sexualorientation, gender, gender identity, disability, or veteran status, immigration status,socio-economic status, age, weight, or pregnancy status is prohibited. ● Additional guidance on hateful content, terrorism, and violent extremism that violates our Community Guidelines is available here. These Terms make clear the extent to which hate speech and terrorism are prohibited. Thisreduces the likelihood of several risks falling with categories identified by the DSA, including inparticular: (i) the dissemination of illegal hate speech and other types of misuse of Snapchat forcriminal offenses and the conduct of illegal activities in Category 1, (ii) the right to humandignity, non-discrimination and child rights in Category 2, (iii) negative effects on public securityin Category 3 and (iv) negative effects on Teens, as well as serious negative consequences toa person’s physical and mental well being in Category 4. We understand that each of the above categories can be nuanced and open to interpretation,that is why we have included explainers for each category. Privacy Policy Snap also publishes a Privacy Policy, which is easily accessible via Search Engine as well aswithin our Privacy Center and in the footer of Snap websites as well as in Apple/Android stores. Itis machine readable and easily understandable. 207 CONFIDENTIAL Our Privacy Policy provides a detailed description of our privacy practices, including anexplanation of how Snap collects and uses personal data and how individuals can control theirinformation. We recognize that policies and terms can be overwhelming documents. That is whyour long standing approach has been to provide additional, bite-sized information on our generalpractices (Privacy Center), our philosophy to privacy (Privacy Principles), what we do with userdata (How we use your information), advertising (Snap and Ads), and specific products (Privacy byProduct). In the introduction to our Privacy Policy, we state the following: “We’ve done our best to write this Privacy Policy in a way that’s easy to understand for allour users and free of difficult language and legal phrases. If you want to review something later on, you can always take a look at our Privacy Center. We designed it to give you easy-to-digest summaries of our privacy practices. For example, our Privacy by Product page gives product-specific information and links to support pages with tips and tricks. Still have questions? Just reach out to us.” Privacy by Product gives users concise and easily understandable information about ourproducts. For example, this webpage provides an overview of our approach to Snaps \& Chats, aswell as hyperlinks to more detailed information on specific aspects. Similarly, there’s a section onSpotlight, Lenses, My AI, Stories, and many more products. In addition to these documents, wealso show in-app notices that provide more information about our products and services. Advertising Policies Snap also publishes Advertising Policies, which outline the terms and conditions for use of Snapadvertising services. These are easily accessible via Search Engine, machine readable, and easilyunderstandable. 208 CONFIDENTIAL Snapchat users who choose to advertise to other users on Snapchat must agree to our Snap Advertising Policies, including an obligation for advertisements to comply with applicable laws and rules in the European Union and each Member State where the update advertisements willrun. 5.2.3 Product Specific Terms In addition to the Terms described in detail above, we also have specific, publicly-available termsand policies that govern the use of additional aspects of Snapchat’s features: Spotlight Snapchat users who choose to contribute content to Spotlight agree to the Snap Spotlight Submission and Revenue Terms, which are made available to all users prior to submitting a video to Spotlight and were last updated in August 2023. Snap also provides users who submit content to Spotlight with clear Spotlight Guidelines, describing the policy, technical, and legal requirements for submissions to Spotlight, as well as reminding users of the Terms (including our Community Guidelines). Discover We have specific publishing agreements with our premium partners that post content onDiscover, such as media organizations and Snap Stars, that require them to abide by our Terms (including our Community Guidelines). 209 CONFIDENTIAL Lenses Snapchat users who choose to develop and submit Lenses for publication on Snapchat via Lens Studio must agree to the Lens Studio Terms. Lenses must comply with our Lens Studio Submission Guidelines, which also remind users of the Terms (including our Community Guidelines). 5.2.4 Other Aspects Oversight and Administration Change Management Snap's Management body (including Legal and Public Policy team stakeholders) review andupdate various Snapchat Terms and Conditions (including Community Guidelines, Terms ofService, Privacy Policies, and Advertising Policies) for additional information that may result inimpact on our risk and mitigation assessments, and to ensure that the Terms/Guidelinesaccurately reflect the contractual relationship and/or other obligations between users of therespective Services and Snap and adhere to applicable legal requirements. Snap's Legal team has a formal process in place to make and track changes to the Terms andConditions and to communicate key changes to stakeholders in a timely manner. Snap’s Terms are regularly reviewed (and updated as needed) by Legal and Policy teamstakeholders to ensure that they accurately reflect the contractual relationship and/or otherobligations between users of the respective Services and Snap and adhere to applicable legalrequirements. Once the document is finalized and approved, it is then localized in all supportedlanguages, including all official languages of the European Union as explicitly required by theDSA. To the extent the changes to the terms, policies, or guidelines are determined to bematerial, Snap will provide users with reasonable advance notice. Snap provides an in-app pop-up to notify recipients of the service of material changes to theTerms and Conditions. Accessing Terms and Conditions 210 CONFIDENTIAL Accessibility Privacy and Safety Hub Snap’s Privacy and Safety Hub was launched last year and combines our Privacy Center, SafetyCenter, and Transparency Center all under one umbrella site. This is where Snap publishes formaltransparency reports. The rationale behind the integration of these three centers is that we believe there is a naturaloverlap between these areas, and that all the information provided in those domains contributeto providing awareness and building trust with our community and other stakeholders, such asparents, teachers, journalists, trusted flaggers, law enforcement, regulators, and NGOs. The top navigation provides Privacy, Safety, and Transparency resources, as well as our latestNews in those areas. In this section, we highlight a number of areas for illustration purposes, and refer to the website for further information. Privacy Center Our Privacy Policy provides a detailed description of our privacy practices, but we recognize thatpolicies and terms can be overwhelming documents. That is why our long standing approach hasbeen to provide additional, bite-sized information on our general practices (Privacy Center), ourphilosophy to privacy (Privacy Principles), what we do with user data (various pages, including thePrivacy Policy), advertising (Snap and Ads), and specific products (Privacy by Product). Safety Center From the Privacy Center, users can easily navigate to the Safety Center, which provides anoverview of our Safety resources, including tips on how to report content, the acknowledgmentthat safety is a shared responsibility, as well information on our Trusted Flagger Program, Safety Advisory Board, Digital Well-Being Index and more. Again, the goal here is to provide easy to navigate and process information. Transparency Center Our Transparency Center provides additional transparency resources to our users and to the public at large, including our Community Guidelines (see Terms section), Transparency Reportsand EU-specific information required under the DSA. On our EU transparency page, we publish EU-specific information required under the DSA, including the number of Average Monthly Active Recipients of our Snapchat app in the EU, and 211 CONFIDENTIAL information about our legal representative in the EU, how EU law enforcement agencies cansubmit requests to snapchat, and the regulatory authorities that regulate us under the DSA. News Page Snap also frequently publishes Privacy and Safety related information on the Hub’s News webpage. The purpose of these news articles is to inform the general public about recentdevelopments on issues relevant to privacy, safety and transparency on Snapchat. Languages Our Terms of Service have been translated into all official languages of the European Union as explicitly required by the Digital Services Act. Readability As outlined in Snap’s 2023 Risk Assessment and Mitigations Report, our European Snapchattercommunity consists of a diverse range of ages and genders. Snapchat services are not primarilydirected at or used by minors. While Snapchat does have a young demographic, only a small 212 CONFIDENTIAL percentage of European Union users fall within the 13-17 age category. The largest age categoryis 18-24. Snap’s Terms have been designed to be a concise summary in clear, easily understandable,unambiguous language, in EU member state languages, in a publicly available, easily accessible,and machine-readable format., including summaries and explainers. This helps all users tounderstand what activity is prohibited on Snapchat and the consequences, which reduces thelikelihood they will engage in illegal or violating activity. In addition, our Privacy and Safety Hub and Support pages as explained in the Transparency partof our Report have also been designed to be user friendly and easily understandable. Forexample, we created our Privacy and Safety Hub, with pages such as our Privacy by Productpage, to give Snapchatters a high-level summary of our privacy and safety practices across eachof our products and features. We also created a video to visualize our privacy practices, and useicons and other best practices as recommended by privacy and safety experts and therecognised Age Appropriate Design Codes. This helps all users to understand how Snapchatworks, what options they may have, how we moderate and enforce our terms and how they canget support. This reduces both the likelihood of illegal or violating activity and the severity ofharm in the event they are exposed to illegal or violating activity despite our limitations. Teensreading our Privacy Center can understand how their data is being processed by Snap and findmore information about relevant privacy settings which reduces the likelihood and severity ofnegative effects on Teens’ data protection rights. 5.2.5 Conclusion Snap provides terms and conditions for the recipients of its services, which incorporate thecontent and meet the language requirements of the DSA. As explained in Section 4, we have concluded that Snap’s terms and conditions, incombination with the other mitigations explained in this Section 5, are reasonable,proportionate and effective mitigation measures for the risks presented by Snapchat’sin-scope services. 5.3 Transparency Snap is focussed on providing users with the right level of information, at the right time. Weunderstand that our community does not always have time to read multi page documents. This iswhy we strive to provide users with bite-sized information that is easy to access and understand,while also giving them an opportunity to review more detailed information where appropriate. Information provided to users can be divided into three categories: ● Information we provide on our website; 213 CONFIDENTIAL ● Information provided in app stores; and ● Information we provide in our application. 5.3.1 Information we provide on our website At Snap we have a number of avenues to provide information to users. The two primary sources of information outside of our application are our Privacy, Safety, and Policy Hub and our Support Center. Privacy, Safety, and Policy Hub Snap’s Privacy and Safety Hub was launched in 2022 and combines our Privacy Center, SafetyCenter and Transparency Center all under one umbrella, and in 2024 we added a dedicatedPolicy Hub to this center. The rationale behind this change is that we believe there is a naturaloverlap between these areas, and that all the information provided in those domains contributeto providing awareness and building trust with our community and other stakeholders, such asparents, teachers, journalists, trusted flaggers, law enforcement, regulators, and NGOs. 214 CONFIDENTIAL The top navigation provides Policy, Privacy, Safety, and Transparency resources, as well as ourlatest News in those areas. In this section, we highlight a number of areas for illustration purposes, and refer to the website for further information. Policy Center We want Snapchat to be a safe and positive experience for everyone who uses our platform orproducts. For this reason, we created rules and policies that explain the rights and responsibilitiesof all members of our community. Our Policy Center provides a central place for our CommunityGuidelines, Advertising Policies, Content Guidelines, and Commercial Content Policy. We havealso reformatted our policies in a way that is easier and more intuitive for users to navigate,replacing our previously long, text-heavy pages with shorter, more digestible segments with clearheadings and organization by subject matter. Privacy Center Our Privacy Policy provides a detailed description of our privacy practices, but we recognize thatpolicies and terms can be overwhelming documents. That is why our long standing approach hasbeen to provide additional, bite-sized information on our general practices (Privacy Center), ourphilosophy to privacy (Privacy Principles), what we do with user data (How we use yourinformation), advertising (Snap and Ads), and specific products (Privacy by Product). In theintroduction to our Privacy Center we state the following: “Privacy policies tend to be pretty long – and pretty confusing. That’s why we did our best to make our Privacy Policy brief, clear, and easy-to-read! 215 CONFIDENTIAL You should read our entire Privacy Policy, but when you only have a few minutes or wantto remember something later on, you can always take a look at this summary – so youcan learn or recall some of the basics in just a few minutes.” Since our 2023 Report, we have made additional updates to our Privacy Policy in February 2024.This update included in particular the following changes: ● The addition of summaries to each section of the Privacy Policy to highlight keytakeaways. ● Provided more detail on how Snapchatters can control their information, who they canshare content with, and who can contact them. ● Added additional transparency on how we collect and use Snapchatter information andcontent, with examples of what data we use for purposes like personalization, advertising,and building machine learning models. Privacy by Product gives users concise and easily understandable information about ourproducts. For example, this webpage provides an overview of our approach to Snaps \& Chats, aswell as hyperlinks to more detailed information on specific aspects. Similarly, there’s a section onSpotlight, Lenses, My AI, Stories, and many more products. 216 CONFIDENTIAL Safety Center Our Safety Center provides an overview of our Safety resources, including tips on how to reportcontent, the acknowledgment that safety is a shared responsibility, as well information on our Trusted Flagger Program, Safety Advisory Board, Digital Well-Being Index and more. Again, the goal here is to provide easy to navigate and process information. Since our 2023 Report, wehave included direct links to our existing pages for the Safety Advisory Board, Digital Well-BeingIndex and Information for Law Enforcement, as well as a new support page on FinancialSextortion, in the navigation bar of the Safety Center. In the near future, we’ll also add a page forthe Council for Digital Well-Being. We have dedicated a page on our Safety Center to reporting. Our community, trusted flaggers, and other stakeholders play a vital role in the safety of our platform. A primary way they do this isby reporting content. That’s why we think it’s crucial to raise as much awareness as possibleabout reporting. The dedicated page summarizes the various ways users can report content, and provides additional resources on how to report (e.g. a hyperlink to our Safety Snapshot episode on reporting). The page also links to our Reporting Quick Guide and contains a hyperlink to our web reporting form. Another important component of the Safety Center is our Safety Resources and Support page. The goal of this page is to provide users with additional resources, such as a hyperlink to MindUp, information about our Here For You tool, and country specific information. Since our 2023 Report, we have also included additional resources including: ● a page dedicated to explaining Financial Sextortion; and ● a page dedicated to sexual risks and harms, in an effort to support those in distress. 217 CONFIDENTIAL MindUp is a non-profit organization that supports children ages 3 to 14 by providing them with thetools and knowledge to manage stress and thrive in school all while maintaining optimism,resilience, and compassion. Our Here for You search tool, which is accessible within the Snapchat app, shows resources fromexpert localized partners when users search for certain topics related to mental health, anxiety,depression, stress, suicidal thoughts, grief and bullying. Our country-specific resources provide users with additional information about resources that areavailable to them in their country, such as children’s helplines, suicide prevention hotlines, andmore. See for example the below, for France: 218 CONFIDENTIAL Parents In September 2023, we launched a dedicated microsite: parents.snapchat.com to provide even more information for parents. We recognize that not all caregivers, parents and teachers useSnapchat. Their lack of familiarity may create questions, and may also make it difficult for them tohave a conversation with younger users. To address this concern, the dedicated micrositecontains: an updated Snapchat 101, a specific page on Safeguarding for Teens, a Tools \&Resources section. This has replaced the ‘Parents and Educators’ section of the Safety Centre. The Snapchat 101 page incorporates our previous ‘Parent’s Guide to Snapchat’ but lays it out in an accessible manner (including a short video). 219 CONFIDENTIAL The Safeguards for Teens page summarises the key protections for teens First, we launched Family Center, our in-app tool for parents and caregivers. To help develop Family Center, we worked with families to understand the needs of both parents and teens,knowing that everyone’s approach to parenting and privacy is different. We also consulted withexperts in online safety and wellbeing to incorporate their feedback and insights. Our goal was tocreate a set of tools designed to reflect the dynamics of real-world relationships and fostercollaboration and trust between parents and teens. In the coming weeks, we will add a newfeature that will allow parents to easily view new friends their teens have added. In addition to Family Center, we also created a Parent’s Guide to Snapchat. The Parent’s Guide helps parents navigate the Snapchat app, outlines Snapchat’s and provides parents withadditional information that empowers them and their family to safely express themselves, live inthe moment, learn about the world, and have fun together. Transparency Center Our Transparency Center provides additional transparency resources to our users and to the public at large, including our Community Guidelines (see Terms section), Transparency Reportsand EU-specific information required under the DSA. On our EU transparency page, we publish EU-specific information required under the DSA, including the number of Average Monthly Active Recipients of our Snapchat app in the EU, and 220 CONFIDENTIAL information about our legal representative in the EU, how EU law enforcement agencies cansubmit requests to snapchat, and the regulatory authorities that regulate us under the DSA. Since 2015, we have also been publishing Transparency Reports twice a year, to provide insightinto Snap’s safety efforts and the nature and volume of content reported on our platform. We arecommitted to continuing to make these reports more comprehensive and informative to the manystakeholders who care deeply about our content moderation and law enforcement practices, aswell as the well-being of our community. As part of our DSA compliance, Snap will be adding newmetrics and information to its Transparency Report. Copies of our most recent and previous Transparency reports can be found on our Transparency Report and Previous Reports webpages. News Page Snap also frequently publishes Privacy and Safety related information on the Hub’s News webpage. The purpose of these news articles is to inform the general public about recentdevelopments on issues relevant to privacy, safety and transparency on Snapchat. For example, a recent article introducing Snap’s Inaugural Council for Digital Well-Being, or an article informing the public on Snap’s approach to keeping its community safe during the 2024 Paris Olympics. 221 CONFIDENTIAL 5.3.2 Information provided in app stores Prior to downloading Snapchat, we provide users with information about the Snapchat app in theApple and Google Play Stores. This includes general information on the functionalities of the app,as well as information on our data collection practices, and links to our website, Privacy Policyand Terms. This way users are able to get a better understanding of the application ahead ofusing Snapchat. 222 CONFIDENTIAL 5.3.3 Information we provide in our application Once users have downloaded Snapchat, they are required to create an account before they canstart using the application. At Snap, our philosophy is to provide timely notifications and generateawareness at points in time where we believe they will be most effective. We provide a high leveloverview of our onboarding process and highlight examples of our “just-in-time” in-appnotifications in this section. Onboarding process Step 1. When users open Snapchat, they are invited to log in (if they have an existing account) or createa new account. The first set of notices users receive relates to notification settings, and the abilityto connect their device’s contacts to find friends. Both steps are optional. The reason we prompt users to turn on notifications is that Snap isprimarily a messaging service and notifications provide an essential utility when using the service.Snap is intended for real friends and family, and requires users to accept friend requests or bealready existing contacts before they can start communicating with each other. Typically, users 223 CONFIDENTIAL already have their close friends and family stored in their device contact book, so the “Find YourFriends” prompt is intended to make it easier for users to send friend requests to other users andto communicate with one another. Step 2. The second step of the onboarding flow requests basic account information such as the user’sfirst name, last name (optional), birthday and username. When asking for their birthday, we show users a neutral age screen, and if a user selects an ageunder 13, they are prevented from creating an account. We don’t notify the users the reason for afailure to create an account. We have drawn on guidance from the UN Convention on the Rights of the Child69 and UKAge-Appropriate Design Code70 to adopt a risk-based approach to age verification in our agegating process. We considered the risks of the platform as well as the rights of younger user’sright to privacy, freedom to access information and freedom of expression under the Conventionand balanced them against safety risks. We believe more invasive age gates come at a privacycost for all users, and also disproportionately impact marginalized groups who may not haveaccess to government IDs.71 We have supported the UK Online Safety Bill amendment to requireApp Stores to play a more active role in sharing age signals to all app stores. We believe this isthe better upstream solution to address any systemic risks associated with underage usersaccessing platforms. If a user has inputted an age of 13 or older, they are prompted to provide a username. We checkusernames against our Abusive Language Detection (ALD) models. If users type in an abusiveusername (i.e., one that does not comport with our Terms), they are prevented from creating anaccount and are asked to enter a username that adheres to our Terms. Step 3. The third step of the onboarding process is focussed on password creation and providing aphone number and / or an email address. These are standard steps to improve account securityand provide Snap the ability to communicate with users. Step 4. Lastly, we offer users the ability to start finding friends on Snapchat, and the option to create aBitmoji. Snapchat shows Bitmojis instead of profile pictures. Bitmojis protect the identity of users,and prevent abuse from predators who may use profile pictures as signals to reach out to theirtarget victims. 71 See for example the report on age verification issued by the Australian eSafety Commissioner, url. 70 UK Information Commissioner’s Office, Introduction to the Children’s code, url. 69 UN OHCHR, Convention on the Rights of the Child, url. 224 CONFIDENTIAL Just-in-time notifications Once a user has created an account, we create awareness at a feature-specific level, typicallyusing Just-in-Time notices or “JITs”. We conduct user research and sentiment studies, andfeedback we receive from users is that JITs or icons are more effective to inform users than longtext. Below we provide some examples of JITs that create feature-specific awareness. Snap Map Snap Map is off by default and user location data is off by default. Users choose who can seetheir location. Since our 2023 Report, we have simplified location sharing on Snap Map. Location sharing stillrequires a two step permission to enable location sharing: 1. Users will choose location permissions in device settings (e.g. for iOS these are “Never”,“Ask Next Time or When I Share”, “While Using the App”, “Always”). 2. Users must also select sharing with “My friends”, “My Friends, Except” or “Only theseFriends”. However, users no longer have to navigate two separate flows for “live sharing” and “last activesharing”. The selected Friends will see “live location" if the user chose “Always” in device settingsor “last active” if the user chose “While Using the App” in device settings. Users can still decide to turn on Ghost Mode at any time when they want to go off the grid. Spotlight Before a user submits a Snap to Spotlight they are presented with a JIT informing them thatSpotlight submissions are public. This is to create awareness that Spotlight is different from MyStory submissions, which can be shared with friends only, unless the user actively chooses toshare them with “Everyone”. 225 CONFIDENTIAL Thematic awareness and notices Across Snapchat, we offer a number of resources to users to raise awareness on safety topicsand protect them. For example: 1. Here for you If a user types in “selfharm” or related terms in our Search functionality, we try toprominently show them relevant ‘Here For You’ resources among the search results. 226 CONFIDENTIAL 2. Safety Search terms like “safety” will direct users to our relevant Here For You resources, such asinformation on our Family Center, and to our Safety Snapshots, our official channel forsafety and privacy tips and tricks. 3. Heads up If a user types in terms related to illicit drugs in our Search functionality, we try toprominently show them our ‘Heads Up’ resources among the search results. Heads up is 227 CONFIDENTIAL our in-app tool that surfaces educational content from experts to Snapchatters if they tryto search for drug-related content. Our expert partners include the Centers for DiseaseControl and Prevention (CDC), the Substance Abuse and Mental Health ServicesAdministration (SAMHSA), Community Anti-Drug Coalitions of America (CADCA),Shatterproof, Truth Initiative, and the SAFE Project. We also run campaigns on Snapchat to raise awareness about certain themes. For example, on Global Data Privacy Day 2024, we informed the general public about our new Privacy Policy, announced the updated parents guide to Snapchat, and launched a dedicated page on Privacy through Security, as relaunched our interactive Lenses with tips on how to stay safe online. Similarly, on Safer Internet Day 2024, we raised awareness around parents’ options to participate and monitor their child’s online activities through Snapchat’s Family Center. We publish updates on efforts and campaigns to raise awareness around the dangers of fentanyl, and continue to partner with organizations like Song For Charlie to combat illicit drugs on Snapchat. 228 CONFIDENTIAL 5.3.4 Languages As explained above, our Terms of Service have been translated into all official languages of the European Union as explicitly required by the Digital Services Act. However, Snapchat itself is onlyavailable in certain official languages of the European Union and not all. As a result, our in-appand publicly accessible information is also only available in certain official languages of theEuropean Union. We consider it reasonable and proportionate and effective to offer ourmitigation measures in the same languages as Snapchat as we anticipate recipients only usingSnapchat if they understand one of the available languages. 5.3.5 Conclusion Snap offers a wide range of in-app and publicly accessible information to raise awareness aroundprivacy, safety and security to its community and external stakeholders. Our approach is thatthese tools should be easily accessible, easy to use and understand, and provided in a timelymanner.We believe that the awareness measures we have in place provide reasonable,proportionate and effective mitigations. As explained in Section 4, we have concluded that Snap’s awareness raising information, incombination with the other mitigations explained in this Section 5, is a reasonable,proportionate and effective mitigation measure for the risks presented by Snapchat’s in-scopeservices. 5.4 Content Moderation 5.4.1 Approach Across Snapchat, we’re committed to advancing safety while respecting the privacy and freedomof expression of our community. We take a balanced, risk-based approach to combating harms —combining transparent content moderation practices, consistent and equitable enforcement, andclear communication to hold ourselves accountable for applying our policies fairly. Safety is a priority across Snapchat, and we use a combination of in-app reporting, automationtools, and human review to combat harms on the platform. All content must adhere to our Terms, including our Community Guidelines and Terms of Service, and some content must also adhere to our Content Guidelines for Recommendation Eligibility. We strive to be transparent and consistent in our practices and enforcement, while striking the right balance between privacy and safety. 229 CONFIDENTIAL Snapchat Design and Function As a reminder, we have also designed Snapchat with privacy and safety in mind, and this designis key in helping to prevent the spread of harmful and illegal content. Snapchat does not offer anopen news feed where unvetted publishers or individuals have an opportunity to broadcast hate,misinformation, or violent content. We think about content on our platform in two categories: 1. “Broadcast content” is recommended for broad distribution on Snapchat. Broadcastcontent includes Spotlight, “Discover” content on the Stories tab, Lenses, andAdvertisements. 2. “Private content” is distributed to friends/followers. Private content includes PrivateStories, Chat, Groups, Accounts. More information about our mitigations relating to Snapchat’s design and function can be found in Section 5.1 (Snapchat Design and Function). Community Guidelines and Terms of Service When considering our Content Moderation approach, it is also important to bear in mind that all content everywhere on Snapchat must adhere to our Community Guidelines and Terms of Service. Then, in order to be eligible for algorithmic recommendation beyond the creator’s friends or followers, content must meet the additional, higher standards described in our Content Guidelines for Recommendation Eligibility. More information about our mitigations relating to Terms and user awareness can be found in Section 5.2 (Terms) and Section 5.3 (Transparency). Content Moderation Our content moderation processes assess each piece of content against the above Terms,policies and guidelines to determine if that content is compliant. Main policy violation categoriesinclude: Sexual Content, Harassment \& Bullying, Threats, Harm \& Violence, Harmful, False \&Deceptive information, Illegal or Regulated Activities, and Hateful Content, Terrorism, and ViolentExtremism. We assess content with proactive moderation using a combination of automated tools and humanreview to moderate content recommended for broad distribution on Snapchat, such as content inSpotlight, Discover, Lenses and Advertisements and non broadcast content such as stories.Technology review and our Machine Learning models are built based on our policies, so policiesare applied consistently across both automoderation and human review. Proactive detectionmechanisms or in-app reports may trigger a review, at which point, our tooling systems processthe request, gather relevant metadata, and route the relevant content to the moderation team viaa structured user interface that is designed to facilitate effective and efficient review operations.Moderators are Trained on Snap’s guidelines, relevant processes, and tooling. For more public information on our moderation practice, see our transparency reports. 230 CONFIDENTIAL More information about our mitigations relating to content moderation can be found in thefollowing paragraphs of this Section 5.4. Enforcement We provide in-app and web-based reporting tools that enable EU users to report content andaccounts they think violate our Terms, which expressly prohibit the dissemination of harmful andillegal content on Snapchat. We also have mechanisms enabling non-users in the EU to reportcontent on Snapchat they believe is illegal. We respond to user reports quickly, and we usefeedback to improve the content experience for all Snapchatters. User reports are classified bythe reporting reason for moderator review. More information about our mitigations relating to enforcement can be found in Section 5.5 (Enforcement). 5.4.2 Content Moderation We detect violations of our Terms through both proactive and reactive moderation. Our proactivemoderation relies on technological tools (e.g., machine learning) as well as human review. Ourreactive moderation processes are triggered when we receive a report of an alleged violation onSnapchat. Our reporting systems provide users and non-users in the EU the ability to easily report Snapchataccounts and content they believe violates our Terms. We review all flagged accounts andcontent against our Terms. When we determine that a user has violated our Terms, we mayremove the offending content, terminate or limit the visibility of the relevant account, and/or notifylaw enforcement. Our policies and systems promote consistent and fair enforcement, and provide Snapchatters anopportunity to meaningfully dispute enforcement outcomes through Notice and Appealsprocesses that safeguard the interests of our community while protecting Snapchatters’ rights.We continually strive to improve our enforcement policies and processes and have made greatstrides in combating harmful and illegal content and activities on Snapchat. Proactive Moderation (Content Reviews) We use a combination of automated tools and human review to proactively moderate broadcastcontent across Snapchat, i.e., content recommended for broad distribution on Snapchat, such ascontent in Spotlight, Discover, Lenses and Advertisements and non broadcast content such asstories. 5.4.3 Conclusion Safety is a priority across Snapchat, and we use a combination of in-app reporting, automation tools, and human review to combat harms on the platform. All content must adhere to our Terms, 231 CONFIDENTIAL including our Community Guidelines and Terms of Service, and some content must also adhere to our Content Guidelines for Recommendation Eligibility. We strive to be transparent and consistent in our practices and enforcement, while striking the right balance between privacy and safety. As explained in Section 4, we have concluded that Snap’s measures to moderate illegal orviolating content, in combination with the other mitigations explained in this Section 5, arereasonable, proportionate and effective mitigation measures for the risks presented bySnapchat’s in-scope services. 5.5 Enforcement 5.5.1 Introduction We strive to continuously update and improve our enforcement mechanisms to protectSnapchatters and our broader communities. As explained in the Terms part of this Report, Snaphas carefully developed its Terms with a view to mitigating the systemic risks it has identified for the EU (see Section 4 of this Report - Risk Assessment Results). Integral to our risk mitigation efforts are Snap’s policies and processes to enforce these Terms. Below, we explain how weenforce our Terms in a transparent, consistent and equitable manner, balancing our commitmentto safety with respect for the privacy interests of our community. 5.5.2 Protections against Misuse (Art. 23) Snap has implemented a Strike Roadmap to perform enforcement actions against users whorepeatedly post illegal content or violate the Terms and Conditions. The enforcement action isdetermined based on the severity of the violation in a given timeframe. Snap's Trust and Safety team has a process to review and update, if necessary, the StrikeRoadmap. 5.5.3 Transparency for Measures of Protection Snap publishes information within its Terms and Conditions information regarding its policies andenforcement mechanisms related to misuse of the platform, including the following: ● Snap’s Community Guidelines related to Illegal Content ● Snap Illegal Content Explainer ● Snapchat Moderation, Enforcement, and Appeals 232 CONFIDENTIAL 5.5.4 Notification of Criminal Offenses (Art. 18) Proactive referrals to law enforcement and governmental agencies As discussed above, we have various tools within the app where individuals may report illegal orharmful activity to our Trust \& Safety team. The Trust \& Safety team then will investigate the reportand if needed, take action, which can range from deleting the content and warning the user tolocking the violating account. In the event the report concerns any threat to a person’s life or safety, Snap will alert theappropriate authorities. Snap’s Content Moderators are trained to preserve relevant information, including useridentifying information, and content, regarding the involvement of law enforcement. Law enforcement takedown requests (Articles 9 and 10) EU Law Enforcement (“LE”) may submit orders for takedown of content or accounts (Article 9) viaemail to a dedicated email address. Similar to the process for takedown requests, EU LE can submit specialized requests to provideinformation (Article 10) via email to a dedicated email address. 5.5.5 Complaint Handling System (Art. 20) Snap provides users to access the internal complaint handling system where they are able toappeal content removals or account takedowns dependent on the relevant policy and type ofcontent violation. Snap's internal complaint handling system is configured to record thesubmission of complaints for a 6 month period. 233 CONFIDENTIAL Statement of Reasons (Art. 17) We provide further detail regarding each step of this reactive moderation process below. Notice to Reporter After a reporter reports a piece of content or an account for violating our Terms or causing harm,we will immediately send them a notice confirming receipt of their report and assuring them thatwe are investigating it. 234 CONFIDENTIAL We endeavor to resolve all reports as quickly as possible while ensuring that we do a thoroughreview and achieve the correct result. Some reports can be resolved much more quickly thanothers, which may be more nuanced and require escalation and/or input of other teams.However, on an average basis, our Trust \& Safety team resolves reports regarding particularpieces of content in two hours, and reports regarding chat and account-level concerns in 24hours. Snap publishes information about the Median Turnaround Time to process appeals in its Transparency Report. 5.5.6 Effectiveness of Enforcement We regularly meet with NGOs and other stakeholders to discuss our measures and generally receive positive feedback. For example, in a meeting in January 2024 with Child Focus, a Belgian Safer Internet Centre operating hotline, helpline and awareness centre, Snap was praised for itsresponsiveness. They stressed that our system works well and we act fast when it comes to 235 CONFIDENTIAL taking down illegal content and escalating and that they are pleased with their collaboration withus. Since our 2023 Report, there are now no illegal or other violating content categories in our lowlikelihood category and we have observed a further decrease in our overall PVP compared to lastyear. See Section 6.4 (Prevalence Testing) for further details. 5.5.7 Conclusion Increases in reporting, enforcement and proactive law enforcement referrals over time do notmean that Snapchat has become less safe. On the contrary, these upward trends correlate with acontinued drop in Policy Violating Prevalence (PVP) on Snapchat overall since our 2023 Report.In other words, as we get better at detecting and enforcing against an increased number ofviolations, the frequency of violations found on Snapchat continues to decrease overall. We are committed to continuously improving the safety of our communities on Snapchat andbeyond, and use prevalence testing to identify and adapt to changing abuse trends on Snapchat,so we are best equipped to detect and address any gaps in enforcement. As explained in Section 4, we have concluded that Snap’s measures to enforce its Terms, incombination with the other mitigations explained in this Section 5, are reasonable,proportionate and effective mitigation measures for the risks identified for Snapchat’sin-scope services. 5.6 Algorithmic Systems 5.6.1 Introduction This section describes specific mitigation measures that Snap has put in place with regards toSnapchat’s algorithmic systems for the in-scope services on Snapchat to address the risksidentified in its risk assessment pursuant to Article 34(1), DSA. In line with Article 34(1), our risk assessment is proportionate to the risks identified taking intoaccount their severity and probability, and the design of our recommender systems and otherrelevant algorithmic systems. While Snapchat uses several algorithmic systems across all ofSnapchat’s in-scope services: Spotlight, Discover, Map, Lenses, Public Profiles, Advertising, therisks identified in our risk assessment focused on the following algorithmic systems: 1. Content Recommendation Systems in Spotlight and Discover 2. Advertising Systems 3. Content Moderation Systems 236 CONFIDENTIAL While we do have mitigation measures in place for other algorithmic systems, such as algorithmicsystems relating to our Lenses and Maps products (as explained, for example, in Section 5.4(Content Moderation)), we have therefore focused our efforts on the above contentrecommendation and advertising systems when considering specific mitigation measurespursuant to Art 35(1); other algorithmic systems are not considered in this Section 5.6. The specific mitigations put in place for our Content Moderation Systems and AdvertisingSystems are covered in Sections 5.4 and 5.7 respectively. This Section 5.6 is therefore primarilyfocused on the specific mitigations put in place for our Content Recommendation Systems inSpotlight and Discover (which we refer to in this Section as the “content recommender systems”). As noted in the What’s New and Introduction sections of this Report, we are consideringsimplifying the Snapchat application from 5 to 3 tabs. Rather than having the current separateSpotlight and Discover tabs, this envisages a single unified tab for more public content to theright of the Camera. This may also result in a unification of our recommendation systems acrossSpotlight and Discover. We plan to run some tests later this year to assess how it performs. This isprimarily a cosmetic change, and it should not impact the mitigations including with respect to thedesign and function of these recommendation systems. For now, we have continued to highlightbelow a few differences that exist between the content recommender systems for Spotlight andDiscover. 5.6.2 Content Recommendation Systems Snap provides a free personalized content experience that is intended to entertain and delightusers in the same app they use to communicate with their friends and family. Users find newcontent on Discover and Spotlight primarily through our algorithmicpersonalization/recommendation service. Algorithmic content recommendation systems, like the one Snapchat uses, present a number ofrisks. For example, they may give rise to, amplify and/or result in the rapid and widedissemination of illegal content and/or other harms identified in Section 4, if not adapted andtested appropriately. We have designed our systems and processes to mitigate these risks. Thisincludes the use of appropriate descriptive terminology, robust automated and humanmoderation, sufficient transparency with our users about the functionality of these systems, theability to opt out of personalization, and the other mitigations as described in the testing andadaptation section of this document. How do our Content Recommender Systems work? To help users discover content they will be interested in, Snap’s content recommender systemsseek to understand the types of content viewers are interested in and not interested in. See 237 CONFIDENTIAL https://help.snapchat.com/hc/en-gb/articles/17338132910484-Personalisation-on-Snapchat for information on how our Recommender Systems work Benefits Snap’s recommender systems allow users to more easily discover interesting, entertaining, andrelevant content. With over a million submissions a day of content, discovery methods like sortingby popularity, alphanumeric, timestamp, or curation are not practical. Our recommender systems help viewers discover new interests they otherwise would have neverfound, and help creators who otherwise would not have been able to find an audience, allowusers to learn, develop, play and have fun online. Users can explore different experiences, learnabout topics of interest, and see what is happening around the world. Recommender systems aredynamic and responsive in that they can respond to viewers feedback. We know users consider personalized recommender systems to provide significant benefitbecause: ● Viewers tell us (through their actions) that they prefer recommendations over otherapproaches and access to entertaining content is one of users’ most frequent requests;and ● When we have tested removing personalization on Snapchat, we see a significant fall inuser engagement (view time). We also note that one of the reasons that traditional media services (i.e. linear television,newspapers, and magazines) are perceived to be in decline is because they are less entertainingto a diverse audience than the personalized alternatives provided by online platforms, such as Snapchat’s in-scope services. 5.6.3 Oversight and Administration Algorithmic System Review Snap conducts a comprehensive review of its Algorithmic Systems. The purpose of this review isto centrally catalog algorithmic systems that are significant to the functioning of Snapchatproducts as well as to safeguarding user safety and fundamental rights. This process is used toconfirm understanding and documentation of significant algorithmic systems and reviewalignment of algorithmic systems with Snap’s standards of care for them. 238 CONFIDENTIAL 5.6.4 Adaption and Testing In line with Article 35(1)(d), we explain in this part of the Report the extent to which we haveadapted and tested our algorithmic recommendation systems to help address the risks identifiedin Section 4 of this Report. Summary Snap has extensively adapted its algorithmic recommendation systems to ensure our contentexperience is beneficial to users, and that the risks of algorithmic personalization are mitigated. Considering each risk and its mitigation(s) in more detail: Illegal or violating content As explained in this Terms section of this Report, all content on Snap must comply with our Termswhich requires all public content on Snapchat to be suitable for users as young as 13, includingour Community Guidelines. Additionally, content personalized by our algorithmicrecommendation system must also comply with our more restrictive Content Guidelines forRecommendation Eligibility. As explained in the Moderation section of this Report, we have adapted our recommendersystems and its processes to enforce our content policies with robust automated and humanmoderation Our restrictive Terms and robust moderation help Snapchat mitigate the risk that illegal, false, orinappropriate content will be available to be promoted by our recommendation algorithms. As explained in the Enforcement section of this Report, users may also easily report inappropriateand illegal content. Each piece of content in Spotlight and Discover has a menu that allows usersto report content. All reported user-generated content in Spotlight, Discover and Ads is reviewedby human moderators. If the content violates our policies and somehow made it through ourautomated and human reviews, it is made ineligible for future recommendations by ouralgorithmic systems. The effectiveness of these measures is tested through prevalence testing and by reviewingprivacy and other consumer queries raised to our community support teams, our Data ProtectionOfficer and our DSA Compliance Team. Lack of user understanding Our recommender systems are complex and the process, the signals used in ranking and howsignificant each signal is to the recommender system can be challenging for users to understand. 239 CONFIDENTIAL To help users and answer frequently asked questions, and as part of our DSA compliance, wehave: 1. Adapted our content to include links to articles available explaining how we personalize content in Spotlight, Discover and Ads here. This includes a description of the main parameters used for our recommender systems, as well as the weighting applied to eachsignal. 2. Users may also reach out to our Support team if they have concerns or questions abouthow our algorithms work. We test this is appropriate by reviewing privacy and otherconsumer queries raised to our community support teams and our Data Protection Officer. Intrusive personalized recommendations We believe content is more relevant and entertaining when it’s personalized to a user’s interests,and not to someone else’s. However, there is a risk that some users may experiencepersonalized recommendations based on their inferred interest to be intrusive. In Discover and Spotlight users can disable personalized content by either tapping on ‘...’ then‘Why am I seeing this content?’ which will take the user to Settings or the user can navigatedirectly to Settings and ‘European Union Controls’. When the user disables personalization, the 240 CONFIDENTIAL Discover and Spotlight experiences will be less personalized, and rely on essentials to determinewhat content to show the user, such as the language the user has set on their phone, their age,and country. Users will still see content, but it will be more random and less relevant to the user’sinterests (as required under Article 38 DSA). If the user wishes to enable personalization again,users can do so either by tapping on the favorite icon (♥) in Discover and Spotlight and thentapping ‘Enable’ in the ‘Show More Personalized Content?’ screen or by going to Settings in‘European Union Controls’. Discrimination Algorithms that process special categories of personal data (as defined in GDPR) on a large scaleare considered high risk and require explicit user content. We have adapted Snapchat’srecommender systems so they do not track or identify special categories of personal data,including for the purpose of recommending content and ads. Rapid spread of illegal or false content \& crisis exposure There is a risk of rapid and widespread illegal or false content on Spotlight and Discover, as wellas exposure to crisis situations and unexpected events like riots. We combat this risk byprohibiting illegal or false content in our Terms of Service and Community Guidelines andallowing users to report violations. More importantly, Spotlight relies on a combination ofautomated and human moderation on all submitted content before any video receives broaddistribution. On Discover, only approved Creators can have their Stories distributed in Discover.Those that are approved have their Stories and the ‘tile’ art moderated. We also monitor reportingand hide rates on both Discover and Spotlight. In recent major events / crisis situations, such asthe Israel - Gaza conflict and the EU elections, these mitigations were shown to be very effectiveas we did not suffer significant illegal or false content on the in-scope services of Snapchat. Filter bubbles Our recommender system algorithms are designed to serve users with content that they will findengaging based on factors that include which categories of content they have previouslywatched. There is a risk therefore that, without safeguards, the algorithm will tag users who viewcontent that may not be harmful on its own as being interested in that content and that repeatedand frequent exposure to that content could be harmful. For example, while one piece of contentrelated to dieting may not be harmful, if a user sees many or frequent videos about dieting, theuser may feel inappropriately pressured to diet or may get a skewed perspective on how peoplemanage their relationship with food. We address this risk in a few ways. Firstly, we take significant steps to prevent and removecontent that may become harmful when viewed frequently on Spotlight or Discover, including as 241 CONFIDENTIAL explained above and in the Terms, Moderation and Enforcement sections of this Report. Secondly, our content categories do not include harmful content categories and so in the unlikelyevent that a user does view harmful content, this will not be used by our recommender systemalgorithm to recommend similar content. Thirdly, in our Discover and Spotlight contentrecommendation systems, we have rules in place to ensure that a particular category of contentwill only be recommended occasionally to a given user. In other words, if a user is interested inmakeup videos, we’ll try to diversify the content by only showing makeup videos occasionally. We evaluate our recommendations to users in terms of the number of categories of content weare introducing to them, while at the same time ensuring we do not overwhelm them with anyparticular type of content. This helps reduce the risk of filter bubbles, since users will be serveddiverse content even if our models show they have a strong interest in certain types of content. Erroneously excluding content There is a risk that our efforts to ensure appropriate content on Snapchat results in some contentthat is appropriate being mistakenly identified and incorrectly moderated. This may create forexample, a risk to users rights to freedom of expression. To combat this ‘over-moderation’, we evaluate and work to improve our automoderation in termsof precision and recall, and currently have very high auto-approval precision for Discover and Spotlight. In addition, as explained in the Enforcement section of this Report, we have additional moderation transparency messages (statements of reason) and a more comprehensive appealsflow for moderated creators and content as part of our efforts to comply with the DSA. Viewers could be watching but not enjoying content There is a risk that the recommendation systems and models we build end up optimizing only forshort-term metrics like engagement (i.e. time spent) in the Snapchat app, rather than in support ofSnap’s mission of “empowering people to express themselves, live in the moment, learn aboutthe world, and have fun together". Our long-term objective when recommending content to userstherefore goes beyond time spent and is focused on whether our users are enjoying themselvesand are entertained and satisfied with their experience. Snap evaluates the effectiveness at achieving this objective in multiple ways, in particularensuring that we evaluate our algorithmic performance using a wide range of factors and notsolely relying on user engagement (i.e. time spent). In addition, Spotlight has been designed not to distribute sensitive (i.e. shocking) content to13-17-year-old users’ Snapchat accounts, which includes non-glorifying discussion of self-harmand suicide content (such discussion is not prohibited on Snapchat but may still be sensitive). Forusers under 18, we will remove all content labeled as sensitive. For users over 18, we will limit itsdistribution. 242 CONFIDENTIAL We evaluate our algorithms across the above dimensions because we believe they are thedrivers to the ultimate outcomes we are attempting to deliver for users: that they be (1) satisfiedwith our experience - which we survey regularly (i.e. quarterly) across all tabs in our app and (2)continue to use it (i.e. user retention). 5.6.5 Change Management From a high level, Change Management over algorithmic systems at Snap is governed by thepreviously described Privacy and Safety by Design Review process. Material updates toalgorithmic systems and material changes to model pipelines, input data, and third party userdata are documented and reviewed. 5.6.6 Monitoring and Quality Assurance Performance Monitoring Snap monitors deployed algorithmic systems for anomalies and issues and establishes alerts tonotify Engineering teams when potential issues arise. These alerts look for relevant spikes oranomalies in statistics. Quality Assurance Snap monitors algorithmic systems related to content moderation for quality and precision on acontinuous basis. Monitoring may include: ● User Reports ● User Hides ● Content removal and user appeals ● Policy Violative Prevalence (PVP) ● Content Rejection Snap uses dashboards to visualize content moderation statistics and allows various users toestablish their own alerts based on spikes in content moderation activity. Snap Engineers mayalso investigate spikes in user reported content or automatically detected violative content toidentify correlation between model deployment to feed back into broader Engineering teams. 5.6.7 Conclusion Users find new content on Snapchat primarily through our algorithmicpersonalization/recommendation service. While algorithmic content recommendation systems,like the one Snapchat uses, present a number of risks, we’ve designed our systems to mitigatethese risks. This includes the use of appropriate terms, robust automated and human moderation, 243 CONFIDENTIAL sufficient transparency with our users about the functionality of these systems, the ability to optout of personalization, and the other mitigations outlined above. As explained in Section 4, we have concluded that our adaptation and testing of Snapchat’salgorithmic systems described above, in combination with the other mitigations explained inthis Section 5, are reasonable, proportionate and effective mitigation measures for the risksidentified. 5.7 Advertising Systems 5.7.1 Introduction Snap relies on online advertising to support its business. Snap recognises that without mitigationsits advertising systems also have a significant risk of giving rise to the concerns referenced inArticle 34 of the Digital Services Act. Snap Advertising is a digital ad product created foradvertisers who would like to easily create and manage ads that target relevant audiences onSnapchat. We process user information about Snapchatters to serve them with ads withinSnapchat that we think they might be interested in. However, advertising systems in general might give rise to, amplify and/or result in the rapid and wide dissemination of illegal content and/or other harms identified in Section 4, if not adapted and tested appropriately. 5.7.2 How do our Advertising Systems Work? An overview of Snap’s ads services can be found here and here. In essence, Snapchat collects data about our users as they register, log in and use Snapchat. As is described in our Privacy Policy, this data is comprised of: ● Information the user provides us ● Information we collect as the user interacts with Snapchat ● Information we collect from third parties Snapchat Ads Manager and its various tools allow advertisers to leverage this data for targeted advertising. Advertisers can use our Audience Insights tools to see the estimated aggregated demographics, including age, as well as locations, interests and device overviews of theirtargeted audience. User-level data is not directly available to advertisers through thesedashboards. Some of Snap’s advertising tools allow advertisers to benefit from Snap’s use of data about theircustomers such as customer personal data provided by our advertisers and data collected fromthird-party services along with our users’ personal data, to provide and improve ad targeting andmeasurement: 244 CONFIDENTIAL ● Snap Custom List Audiences \- An advertiser and/or their agent can use this service to upload customer list data to Snap via Ads Manager. See the Custom List Audiences section of our Business Help Center. Customer list data provided by advertisers is used tocreate an ‘audience’ of Snapchatters matching the information in the customer list data.This allows advertisers to target ads to that audience, or similar audiences, on Snapchat. See the Custom Audiences Overview in our Business Help Center. ● Snap Pixel and Conversion API \- An advertiser and/or their agent can also use this service to help target their ads on Snapchat: ○ For Pixel, advertisers install a piece of JavaScript within their web pages whichsends data to Snap when those pages are accessed by website visitors. See the Install Snap Pixel section of our Business Help Center. ○ For Conversion API, advertisers install Snap API code on their servers thatfacilitates passing web, app and offline events directly to Snap via Server-to-Server integration. See the Conversions API section of our Business Help Center. ● Advanced and Estimated Conversion are examples of the additional services that we offer to advertisers to target and measure the performance of their advertising using advancedprivacy enhancing techniques. Snap acts as a data processor of data relating to EU data subjects received from advertisers viathe Custom List Audiences, Pixel and Conversion API services. It processes the information inaccordance with advertiser instructions subject to its data processing agreement (which followsrequirements set out in Article 28 of the General Data Protection Regulation (GDPR)). Our ad ranking algorithm determines which ads are displayed to a Snapchatter who is in theselected audience for those ads. The ad ranking algorithm uses various signals, including prior adinteractions and social signals, to determine which ads that user is more likely to interact with andthen combines this with the results of advertiser ad action for that Snapchatter, to select an ad todisplay. Snap analyzes prior ad interactions to target advertisements. For example, we maydetermine that a user is likely to swipe up on certain types of ads or download certain types ofgames when they see an ad on Snapchat. We may then use this information to show that user similar ads. This is explained on our Snap and Ads Privacy and Transparency page. Snapchatter interactions with the ad (i.e. impression data) is then logged to (a) attributeimpressions to conversion events (such as a purchase on an advertiser website or download ofan advertiser app) to demonstrate the performance of the ad and (b) to further train the adranking algorithm. 245 CONFIDENTIAL 5.7.3 Benefits Snapchat is used by millions of people in the European Union. They use Snapchat because itfosters fast and authentic communication with those who matter most to them. It is why ourcommunity continues to grow. We consider it is in the best interest of all our users, including 13-17s, for them to have access tothe best, most entertaining version of Snapchat possible, allowing them to exercise their digitalrights (such as access to information, association with others, have a voice and to play and havefun) regardless of their financial background and ability to pay. We receive feedback everydayfrom our users; calling for new features, functionality and improvements. We are only able to dothis by raising revenue from other sources. In common with many others in the industry, this hasmeant turning to advertising. Our ability to raise revenue by selling targeted advertising opportunities to advertisers meansthat: ● Snapchat is maintained and improved for the benefit of Snap and all recipients regardlessof their ability to pay. If Snapchat was only available for a fee, it would only be accessibleto those who could afford to pay the fee, restricting access to Snapchat and raising risksto fundamental EU rights to information and to access to services, particularly for Teens. ● Snapchatters benefit from being able to exercise digital rights and association with othersonline through Snapchat regardless of their financial background. This includesdeveloping their voice, having fun and access to entertainment and play. Balanced use oftheir personal data also benefits Snapchatters by avoiding seeing advertisements that arenot relevant to them (which is one complaint we have received in the past). AlthoughSnapchatters are given options to manually hide advertisements, through the use ofpersonal data, Snapchatters benefit from targeted advertising by seeing more relevant,age and interest appropriate adverts72. The greater the revenue Snap is able to generatethe more resources Snap can dedicate to supporting access to the service and teens’development. ● Advertisers benefit from being able to promote their brand and products to a Snapchatteraudience most likely to be interested in them. This allows advertisers to focus theiradvertising and avoid spending on the display of advertisements to audiences that arenot likely to be interested. Snap Ads also allows advertisers to better measure thesuccess of their digital marketing campaigns so their quality can be continuouslyimproved. Advertisers are also conscious about safety on Snapchat. With this in mind, inMarch 2024 IAS, we announced a partnership with a leading global media measurementand optimization platform73, to: 73 https://forbusiness.snapchat.com/blog/snap-partners-with-integral-ad-science-brand-safety. 72 N. Fourberg e.a., on ‘Online advertising: the impact of targeted advertising on advertisers, marketaccess and consumer choice’, 2021, url. 246 CONFIDENTIAL ○ Conduct a measurement sample study on the advertiser suitability of our publiccontent, specifically Spotlight and Creator Stories. In the study results, IAS foundthat both Spotlight and Creator content on Snapchat is 99% brand safe.74 ○ Jointly develop a new brand safety reporting solution that would give advertiserstransparency into the percentage of safe and suitable content their ads areappearing against. The new solution launched in June 2024.75 However, notwithstanding the benefits advertising systems bring to our users, to Snap and ouradvertisers, we recognise that our targeted advertising will only operate in the best interests of allour users provided that the processing of individuals’ personal data (including by way of profiling)to facilitate the sale of ads that fund Snapchat does not result in our users being subject to‘economic exploitation’. Privacy and Safety are central to Snapchat’s values. When we firstintroduced advertising to Snapchat, we ensured those advertising systems appropriatelybalanced the legitimate benefits explained above with individuals’ fundamental rights andfreedoms, in line with Snap’s strong privacy and safety principles. We have continued to upholdthese values throughout Snapchat’s life, adapting and testing our advertising systems to mitigate risks they may give rise to as identified in Section 4 of this Report. 5.7.4 Adaptation and Testing In line with Article 35.1.(e), Snap has adapted Snapchat’s advertising systems and adoptedtargeted measures aimed at mitigating the risks presented by its advertising systems, includingby limiting or adjusting the presentation of advertisements on Snapchat, to help address the risks identified in Section 4. Considering each risk and its mitigation(s) in more detail: Reasonable and Proportionate Targeting We recognise that, as a platform, we have a responsibility to raise revenue in an appropriatemanner, and we take this responsibility very seriously. We want to ensure advertisers are nottargeting specific individuals on our platform and that users do not feel like their privacy is beingcompromised by our advertising. We also want to prevent advertisers from manipulating smallaudiences with microtargeted campaigns, particularly for political ads. In order to mitigate this risk: 75 https://forbusiness.snapchat.com/blog/snap-ias-solution-ga. 74 “Brand Safety” is based on the GARM standard, only considering content classified at the “Floor”risk-levels . “Creator content” is image and video user-generated content posted to Public Stories; IASsampled a wide variety of creators in the U.S. to ensure a representative sample. Spotlight content isuser-generated video content that appears in the Spotlight tab on Snapchat; IAS audited content from US,CA, GB, UK, NZ, AU markets. IAS audited both Spotlight and Creator content from Oct 30, 2023 - Jan 2,2024. 247 CONFIDENTIAL ● Most of the ads on Snapchat, including all political ads, require a specific minimumaudience to be targeted. This prevents adverts from being micro targeted. ● Snap generally has a short retention period for user content. We do not store content forexcessive periods solely for monetisation purposes. ● Advertisers can only use our data for ads targeting indirectly via the targeting toolsavailable on Snapchat. Amongst other things, this allows advertisers to target audiencesbased on a limited number of high level interest-based lifestyle categories (SLCs)audiences (none of which are available for targeting 13-17 year olds in EU, UK, Norway andSwitzerland), which we have inferred a Snapchatter may be interested in. They are basedon high level, non-sensitive categories inferences, such as Business News Watchers,Sports Fans, and Fashion \& Style Gurus, that users can see and control in the app, as detailed in this support page. The interest categories are intentionally short-lived (13 months), sufficient to allow a year-on-year comparison. All users can manage theiradvertising interest categories in settings and view them via our Download My Data tool(DMD). None of these SLCs are aimed at 13-17s specifically and the user-level targetingdata is not directly available to advertisers. We feel confident that our approach to advertising is reasonable and proportionate, as we have alow incidence of issues in relation to age targeting. Our approach to targeting minimums is basedon mathematical analysis by our privacy engineering teams. Advertising Policies As explained in Section 4 above, our advertisers could use our advertising systems to disseminate information that is illegal or could otherwise harm users, impact their fundamental EUrights or negatively impact public security or health. As explained in the Terms section of this Report, we ensure advertisers are clear about their obligations, we have robust ad policies to prevent inappropriate and illegal advertising on our platform. The systems used by advertisers to create and submit advertising (such as our SnapAds Manager), have been adapted to require agreement to these Terms and provide easy accessto guidance on what is required. We test advertisers’ compliance with these Terms using our Advertising Review process beforeadvertising can be published. See below for more information. Advertising Review Notwithstanding that advertisers agree to our Terms, they may still deliberately or mistakenlyseek to publish advertisements that violate our advertising policies or the law. 248 CONFIDENTIAL As explained in the Moderation section of this Report, in particular the part relating to advertising moderation, we use a combination of automated and human review to prevent ads that violateour policies or the law from appearing on Snapchat. We reject hundreds of thousands of advertsglobally each month. We have a global team that supports ad moderation across 15+ languagesand is composed of both full time employees (FTEs) and contractors. Ad Review team membersare responsible for reviewing ad submissions to ensure ads abide by Snap’s creative policies and technical requirements. Ad Review team members use Snap’s Advertising Policies to assess compliance. Ads must comply with Snap’s Community Guidelines and Advertising Policies inorder to be approved. Grey area ads are discussed with Snap’s Legal and Policy teams.Depending on the seniority, members of the Ad Review team also collaborate with the Salesteam to create a consistent review experience for our Snapchat partners. Fraudulent advertising accounts for the majority of these rejections and our advertising reviewteams are particularly vigilant for this form of violating advertising. This also includes ensuringinappropriate ads are not targeted at Teens. Our review takes account of the targeted audiencei.e. if the ad is for alcohol and the selected demographic for the ad includes Teens, then it will berejected. We use inferred age, as well as declared age, to help ensure Teen users see ads thatare appropriate for their age. Inferred age is regularly checked to ensure it is up-to-date. We monitor ad reporting and enforcement data to ensure our review process is catching areasonable and proportionate level of violating adverts. We aim to ensure that all advertising review is maintained within a 24 hour SLA window fromwhen the advertisement is created by the advertiser. More information on Snap’s ad review process, including the timeliness of review, are located on Snapchat’s For Business website. Advertising Reporting Although we have an advertising review process in place to prevent the publication ofadvertisements with information that violates the law or our policies, it is possible that some ofthese advertisements may be missed or incorrectly reviewed and be published. As explained in the Enforcement section of this Report, our advertising systems have been adapted with an easy mechanism for advertisements to be reported by Snapchatters from withinthe Snapchat app as being inappropriate along with the reason for the violation. Based on thenumber or reports, we will take down the ad or send it to human review for additionalmoderation. All ads that are reported are reviewed by our human moderation team. Upon reporting the ad,Snapchatters are able to select a reason and write in comments. Both the reporting reason andthe comment are provided in the moderation task, as well as the number of reports. We closelymonitor sentiments of the ads on our platform and when ads are taken down, we inform the 249 CONFIDENTIAL advertiser. We also monitor the aggregate number of reports for advertisements on a regularbasis. We monitor ad reporting and enforcement data. Ad Markers If users are not aware when content is an ad or sponsored or other commercial content, there is arisk that without additional mitigations that this may lead to confusion, deception and exploitation. We automatically place an “Ad” marker on all paid ads that run on Snapchat. Sponsored Lensessay “Sponsored”. Our commercial content policy requires all organic content posted byinfluencers to be marked appropriately. We now offer a “Paid Partnership” tag tool thatinfluencers and users may use when they post commercial content to help them comply with thispolicy and their legal obligations. Ad marker example Sponsored Lens example 250 CONFIDENTIAL See below for examples of the “Paid Partnership” tag tool that influencers and users may usewhen they post commercial content to help them comply with this policy and their legalobligations. Add Paid Partnership Paid Partnership label Paid Partnership Explainer 251 CONFIDENTIAL Transparency and Control Some users may have specific vulnerabilities or other reasons to be concerned about any use oftheir personal data for targeting ads. If users do not understand how advertising works, they maynot be able to confirm whether they should be concerned or exercise any choices they may have. As explained in the Transparency section of the Report, our privacy center provides extensive information regarding our processing of personal information. This includes a dedicated page explaining how we use personal data for advertising purposes. We offer choices for users tocontrol the data that’s used to determine the ads they see. In the European Union, we haveintroduced controls to turn off most personalized ads except those based on real time location,language, age and device type, and this is always turned off for teen users in the European Unionand UK, Norway and Switzerland. All users can restrict our use of third party data and beingincluded in advertiser supplied audience matches for ads targeting. We use pre launch testing and our ad review process to help ensure these controls work asdesigned. 252 CONFIDENTIAL Ads Gallery There is a higher risk that advertising will violate our terms or the law, in particular contentmisleading information, if the Snapchatter community and wider society does not have visibilityinto the history of ads over the past year that have run on Snapchat and some details about thetargeting and reach of those ads. Snap has an ads library (as required under Article 39 DSA) which provides increased transparency for ads - not just political - that are currently running, and historically have run in thepast year, directed to EU users on Snapchat. This ads library is available to anyone, can besearched / filtered / sorted based on pre-defined parameters (e.g. country targeted, advertisername, etc) and includes an API interface as well. This allows anyone to check who has paid for anadvert and, if different, on whose behalf is the advertisement being published. In the CommercialContent section of the Ads Gallery, we also include links to all live organic content that has beenmarked with the “Paid Partnership” tag. Information included for each ad is shown in the screenshots below. When a user clicks on the“See Details” link they are taken to the Ad Details modal on the right. Per DSA guidelines, dataincludes: Main Ad Modal ● Ad Publisher - the natural or legal person who paid for the advertisement ● Brand Advertised - the natural or legal person on whose behalf the advertisement ispresented ● Ad Start Date and Ad End Date - the period during which the advertisement waspresented ● Ad Creative - the content of the advertisement, including the name of the product, serviceor brand and the subject matter of the advertisement ● Total Impressions - the total number of recipients the service reached Ad Details Modal ● Impressions by Member State - aggregate numbers for the recipients reached by country(if 0 recipients were reached, the ad will not appear). ● Targeted devices and demographics - whether the advertisement was intended to bepresented specifically to one or more particular groups of recipients, specifically, devicesand demographics; These options do not support exclusion targeting. Ads Gallery - Ads Data 253 CONFIDENTIAL Per the Advertising Review section above, we use a combination of automated and humanreview to prevent ads that violate our policies or the law from appearing on Snapchat. Ads thatwere delivered and subsequently taken down are marked as Rejected in the Ads Gallery. The Snap Ads Gallery is maintained by the Ads API and Ads Manager teams. The ads library 254 CONFIDENTIAL underwent pre launch testing to ensure it met design specs and will continue to develop basedon further testing. Freedom of Expression The purpose of Snap Ads is to amplify advertisers’ commercial messages, and as a result thecontent is rarely political or rather than expressing views. We have specific procedures forpolitical ads. As a result, the risk of a negative impact on freedom of expression from Snap’sother mitigations listed above is low. 5.7.5 Conclusion Targeted advertising on Snapchat is necessary to ensure we can continue to provide a freeservice to all users regardless of their ability to pay. We have taken extensive steps to ensure ourapproach to targeted advertising appropriately balances the interests of Snapchatters, Snap andadvertisers. We have also put in significant measures to prevent fraudulent and other advertisingthat may be harmful or against the law. We reject thousands of them each month to keepSnapchat’s community safe. As explained in Section 4, we have concluded that our adaptation of Snapchat’s advertisingsystems described above, in combination with the other mitigations explained in this Section5, are reasonable, proportionate and effective mitigation measures for the risks identified forSnapchat’s in-scope services. 5.8 Protection of Minors 5.8.1 Introduction Snap’s utmost priority is the safety and wellbeing of our users, especially young people, aiming toensure they have positive experiences on our service. Privacy, safety, and security are key valuesof the company and sit at the core of our value proposition to our users. Since Snapchat’s inception, we have embraced a privacy and safety by design approach and recognise that our platform architecture and product choices play a major role in the protection of our users,including minors. We put significant thought and consideration to ensure our values are reflectedin the architecture of our platform, and in the design and implementation of our products,policies, and enforcement actions. We take the protection of Teens seriously on Snapchat. We’ve designed Snapchat to protect theirprivacy, safety, and security. Our key tenets include acting in the best interests of Teens, offeringstrict default settings for all users, and respecting Teens’ freedom to express themselves safely,while recognizing their right to information about the world. We aim to achieve these tenets bypositioning parents and guardians to help guide teens in their responsible use of our platform, 255 CONFIDENTIAL attaching a heightened safety interest to Teens using our products, and establishing processes toensure we develop products in a way that upholds these tenets. We have implemented thesetenets through the use of Family Center, focusing on age-appropriate content, reporting andblocking mechanisms, and putting in place appropriate protections and limitations on privatemessaging, friending, public content, and advertising. 5.8.2 Overview and Approach Age Appropriate Design Code Snap has adopted the Age Appropriate Design code (AADC) guidance for developers toincorporate appropriate and proportionate measures to safeguard the privacy, safety, andsecurity of users aged 13-17 within platform products and features. This design code includes 15 core standards: 1. Best interests of the child: The best interests of the child should be a primary consideration when we design and develop products. 2. Data protection impact assessments: Product must be covered when appropriate by minor data protection impact assessment. 3. Age appropriate application: Requires a risk-based approach to recognising the age of individual users to ensure we effectively apply the AADC standards. 4. Transparency: The privacy information we provide to Snapchatters, such as our privacy center, support pages and in-app notices, must be concise, prominent and in clearlanguage suited to the age of the child. 5. Detrimental use of data: We should not use children’s personal data in ways that have been shown to be detrimental to their wellbeing, or that go against industry codes ofpractice, other regulatory provisions or government advice. 6. Policies and community standards: Uphold our own published terms, policies and community standards (including but not limited to privacy policies, age restriction,behavior rules and content policies). 7. Default settings: Settings must be ‘high privacy’ by default (unless wecan demonstrate a compelling reason for a different default setting, taking account of the best interests ofthe child). 8. Data minimisation: Collect and retain only the minimum amount of personal data we need to provide the elements of our service in which a child is actively and knowinglyengaged. Give children separate choices over which elements they wish to activate. 9. Data sharing: Do not disclose children’s data unless we can demonstrate a compelling reason to do so, taking account of the best interests of the child.. 10. Geolocation: Switch geolocation options off by default. 11. Parental controls: If providing parental controls, give the child age appropriate information about this. If our online service allows a parent or carer to monitor their child’s 256 CONFIDENTIAL online activity or track their location, provide an obvious sign to the child when they arebeing monitored. 12. Profiling: Switch options which use profiling ‘off’ by default unless there are appropriate measures in place to protect the child from any harmful effects. 13. Nudge techniques: Be mindful of and avoid using nudge techniques to lead or encourage children to provide unnecessary personal data or weaken or turn off theirprivacy protections. 14. Connected toys and devices: If providing a connected toy or device ensure we include effective tools to enable conformance to the Code. 15. Online tools: Provide prominent and accessible tools to help children exercise their data protection rights and report concerns. We continue to actively support the efforts of the Commission and others to establish an EU-wideAge Appropriate Design Code and/or guidance on the application of Article 28 by onlineplatforms, and consider whether further mitigation measures may be reasonable, proportionateand effective for online platforms, ‘gateways’ and other online services. We have recentlysubmitted our feedback to the Commission on its initial proposals for Article 28 guidance andhope to continue to work closely with the Commission and other stakeholders on its adoption.This is also discussed further in Section 5.12 (Codes). Privacy, Safety, and Security of Minors on Snapchat Privacy, safety and security are key values of the company and at the core of our value proposition to our users. Snap has dedicated extensive resources to ensuring protections tosafeguard the rights of Teens on the platform, greatly reducing the likelihood of rightsinfringement. At the highest level Snap follows Age Appropriate Design Codes (or similar)established by the United Kingdom, France, California, etc, as well as our own key tenetsdescribed earlier in this playbook. Snap's approach to privacy and safety by design means thatwe generally design for our youngest users first and work upwards from there. This means ourfirst layer of protection for minors are the mitigations that are designed for Teens but apply to allto advance safety across our community. Snap has put in place a range of mitigation measures to protect the privacy, safety, and securityof minors on Snapchat. This includes general platform safeguards such as our Teen friendly termsand support pages, our moderation and enforcement processes, our parental tools–FamilyCenter, in-app reporting, and Teen specific content moderation and restrictions. In addition to these safeguards for Teens, we consistently enforce our policies disallowing users under the age of 13 from creating or maintaining an account. Persons who are younger than 13 will be blocked from creating an account at the time of registration; accounts that are discovered to be operated by persons under the age of 13 are removed from the platform at the time that Snapchat discovers such violations. 257 CONFIDENTIAL Advertisements for Minors Snapchat automatically disables advertisements based on profiling for users aged 13-17 in the EU.Additionally, safeguards have been put in place to help Teens understand and recognize Lensesand ensure that advertisers and advertisements on our platform comply with our requirements. Snap has implemented additional safeguards and protections for minors related toadvertisements, as described in details in the product-specific subsections below. Identifying Minors As the creators of a central communications tool for young people, we take seriously ourresponsibility to protect teens on our platform. We know age verification is an industry-widechallenge everyone is trying to solve, and we are already working with industry peers, regulators,and third-party technology providers on possible approaches. We look forward to continuingthese productive conversations to achieve methods that work for everyone. Snap currently takes a risk-based approach to provide an age-appropriate experience acrossSnapchat, consistent with best practices such as the UK Age Appropriate Design Code (AADC).As explained in our response to the Commission’s RFI on minors in December 2023: Registration and access to Snapchat In order to download the Snapchat app, users first need to create an account with either Apple orGoogle to access their app stores (Apple App Store and Google Play Store). Both the Apple AppStore and Google Play Store have age restrictions, they require users to create an account beforethey can access the stores, and the age restriction for those accounts is 13+ and in some cases 14, 15 or even 16+ (see Apple and Google age restriction terms). Both Apple and Google rely on declared age to determine if a user is 13+. If a user provides anage under 13 account creation is persistently blocked unless parental approval is provided. Both Apple and Google offer state of the art and easy to use parent tools (see Apple and Google family link terms). This means that in order to download an app - for example Snapchat - from theApple or Google Play Store a user needs to declare to be 13+ or parental approval has beenprovided. Although Snap has asked for access to Apple and Google’s parent tools and age signal to ensureconsistency, increase visibility of our guides and settings, reduce the burden on end users andensure a level playing field with Apple and Google’s own apps, Snap does not currently havesuch access. 258 CONFIDENTIAL As a result, Snap independently asks the user to confirm their age as an additional age assurancemeasure, despite age already being provided by the same user as part of the Apple and Googleaccount registration flow, as follows: Declared age to limit access to Snapchat to its target 13+ audience: a. Our declared age process has been designed to meet industry standards. b. In our Terms of Service, Privacy Policy, and other documentation, we make clear that Snapchat is intended for users 13 years old or older. Users must affirmatively add their birthdate when registering for an account, and we deny users declaring they are under the age of 13 the ability to create accounts. c. If we determine, or are otherwise made aware through an in-app report from a user, parent, or law enforcement, that an account belongs to someone younger than 13, we take immediate action to prioritize and respond to the information. Our trained internal team will review and terminate the account, including immediately deleting the data associated with the account. In respect of a) ● We do not use inferred age techniques to prevent individuals under the age of 13 fromregistering or accessing the app. Reliable age inference is not feasible without data basedon user activity once registered and engaging in the app. We do not have such activitylevel data for any new user at registration, nor do we have this data at any time for usersunder the age of 13 (since all Snapchatters are declared to be 13 or older). ● As explained below, we have stronger age assurance in place to protect minors fromcertain content and features of Snapchat targeted at more mature audiences which allowus to protect against potential U18 use of adult accounts despite not having absoluteknowledge of U13 use. Our approach is stronger as it allows U18 protections to be appliedif there are changes after registration, for example, in potential situations where a parentor other adult user may register an 18+ account using their own information but thenprovide account access and usage to an U18 bypassing any static age assurance appliedon registration. ● Further, as shown in Section 1 (Snapchat Community) the vast majority of Snapchat usersaccess the app to use our messaging services to communicate with friends - not toodissimilar from traditional SMS or other messaging services, which typically do not haveany age gates at all. Such interpersonal communication services fall outside the scope ofthe DSA. In respect of c) and our trained internal team: 259 CONFIDENTIAL ● The team is trained to prioritize these tickets. When our privacy operations team receivesa ticket they act upon this promptly, often within a couple of hours. ● If our team is made aware of an account belonging to someone potentially under 13through external sources (eg. through in-app reporting, Law Enforcement requests), aticket is created and routed to our human Trust \& Safety team who is also trained toprioritize these types of tickets. The team of moderators responsible for reviewing thesereports consists of several dozen FTEs.76 We do not track the specific response time forthis type of ticket, these metrics are tracked across all reporting types. ● We provide moderators with training sessions on policies, processes, tooling, currentevents and cultural norms to be effective at their work. Our moderators are trainedthrough small group training classes and also review a multitude of scenarios whileshadowing high-performing peer moderators. Through practice and instruction, theyapply our policies and enforcement measures in a manner that protects our Snapchatcommunity. This training is conducted over a multi-week period, in which the moderator iseducated on Snap’s policies, tools, and escalations procedures. After the training,moderators must pass a certification exam before being permitted to process content. ● In addition to the general moderation training (see below), these team members receivespecific training and guidance from our (privacy) legal teams. In respect of c) and the use of keywords to detect under age users: ● In general, Snap does not scan keywords during account creation or from accountinformation to detect whether a user might be underage. As noted in Section 1 (SnapchatCommunity), more than 80% of time spent is on private surfaces. Snapchat is primarilyintended and used as a communications tool by our users with their close friends andfamily, and fundamentally different from traditional social media where the majority of thecontent is public. We apply a privacy-first approach to user communications, such asChats, and those are not subject to scanning for purposes of learning a user’s age orprofile. We assessed that doing so would be contrary to the fundamental privacy rights ofindividuals, as well as privacy laws, including the GDPR and ePrivacy Directive. ● However, since our 2023 Report and the RFI, we have begun testing an additionalmechanism to detect under age registrations and access to Snapchat. This involvesscanning text fields within user public profiles for certain key phrases. We assessed the scanning of public profiles to be less intrusive as this content is public,and on balance would not infringe users’ fundamental privacy rights. Where clearstatements are made that the user is under age, the corresponding account is further 76 For a more detailed breakdown of our human moderators please see Section 5.4 (Content Moderation). 260 CONFIDENTIAL investigated and appropriate enforcement action taken (for example, to remove theaccount). We are continuing to test this new mechanism. Access to certain content / features We rely on a combination of declared age and inferred age techniques for stronger ageassurance to limit under 18 access to certain content and features targeted at more matureaudiences. 2. A combination of declared- and inferred-age techniques for stronger age assurance to protect minors. We not only rely on a users' self-reported age, but have techniques to infer a users’ actual age, which considers a combination of various influential signals to limit under-18 access to certain content and features targeted at more mature audiences, including: a. Discover (Publisher Partners): At the time of onboarding, Snap provides its Commercial Content Policy to Publisher partners. Publishers must adhere to Snap’s Commercial Content Policy. This Policy requires Publishers to ensure their content is appropriate for a 13+ audience. If it’s not, our policy requires publishers to age-gate their content. Publisher partners are also given more detailed guidelines on how to comply with the policy (including with respect to age-gating). In addition to this policy, as explained in Section 5.4 (Content Moderation) our global moderation team reviews Publisher Tiles for compliance to ensure content shown in “Stories/Discover” meets our Content Guidelines for Recommendation Eligibility (our standard for “appropriate content”). Our Partnerships team also performs periodic checks of Publishers and Content Creators to ensure that the content they are posting is compliant with these guidelines. Publisher partners receive notification (either emailed or in-app, depending on the creator type) if their content is found in violation of our guidelines. Publisher partners who violate frequently and/or severely are further penalized, after editorial review. These penalties can involve a period of restricted visibility, a suspension during which all publishing is banned, or a permanent channel ban. In rare instances, we've ended relationships with entire organizations. In addition to penalties, Snap regularly communicates updates and clarifications to our guidelines in emailed newsletters to partners, and holds periodic training seminars for publishing partners. b. Spotlight and Discover (UGC): i. Content that is prohibited by our Community Guidelines is prohibited everywhere on Snapchat and our Content Guidelines for Recommendation Eligibility specifies additional categories of content that, while permitted on Snapchat, will not be eligible for recommendation to a wider audience on Spotlight and Discover. These mitigation measures apply to all Snapchatters. 261 CONFIDENTIAL ii. In addition, Spotlight and Discover have been designed not to distribute sexually suggestive content to 13 - 17 Snapchat accounts and only recommends sexually suggestive content to a 18+ Snapchat account if that content has been created by a creator that the account has subscribed to or favorites suggestive content. This uses our combined declared and inferred age techniques. iii. Regarding the machine learning classifiers, we use in-house classifiers that were trained to scan and identify sexually suggestive content using state-of-the-art computer vision models. When user-generated content on Discover and Spotlight is scanned by our machine learning classifiers, content that is scored above our threshold and considered “suggestive” is then removed from the content recommended to teens. Snap assessed (and continues to assess) the effectiveness of our machine learning classifiers via (i) quality testing and product/engineering review before deployment and (ii) ongoing review against in-house human labelling of publicly available content. We do not deploy new machine learning classifiers until they achieve at least 80% precision. Note that we also conduct routine quality checks of our human review where our precision is 95%+. iv. Our systems are designed in a way that a Snapchatter (who is over 18 years old) who has subscribed to a content creator that has posted suggestive content will not see more than one sexually suggestive content video out of seven in their Spotlight feed. If a user hides a video labelled as sexually suggestive or a sexually suggestive creator (here is an example of a piece of Spotlight content that was marked as suggestive), we stop showing that type of content to that user. If a user hides a creator, we stop showing them that creator. c. Lenses: We age-gate certain Lenses (e.g., related to alcohol, gambling, NFTs, etc.). For example, a Vivino Lens will only be shown to 18+ users in the EU, mitigating the risk of such content being shown to users who are under the legal drinking age. d. Ads: We restrict ads based on the user’s age. For example, ads for dating services must be targeted to users over 18 and must not be provocative, overtly sexual in nature, or reference transactional companionship. Similarly, ads for alcohol products must be age-targeted to at least 18+, or the applicable minimum drinking age in the respective country where the ad is running. Our age inference model is used as an integral part of our age assurance method to limit under18 access to content and features targeted and suited to more mature audiences. The inferredage model on Snap uses a variety of influential signals, rather than only relying on the age thatusers provide when signing up to the platform. This helps, for example, prevent regulated ads 262 CONFIDENTIAL from being served to those users who have declared themselves to be over the age restrictionbut we have modelled as likely to be under the appropriate age for such regulated content. For example, if a user’s self-declared age is 20 years old, yet the signals derived from the user’sactivity within the app and the ages of their friends strongly indicate that they are likely under 18years old, we may internally “override” their supplied age and flag the user as a minor, andtherefore filter regulated ad content (e.g. alcohol) from being displayed to them. The inferred age model does not utilise any third party age assurance providers. To minimisedisclosure of personal data, the model is processed within Snap. As flagged above, we do not use inferred-age techniques to prevent individuals under the age of13 from registering or accessing the app. Reliable age inference is not feasible without databased on user activity once registered and engaging in the app. We do not have suchactivity-level data for any new user at registration, nor do we have this data at any time for usersunder the age of 13 (since all Snapchatters are declared to be 13 or older). The use case for ourage inference model has always been geared toward mitigating the risks for users under 18, forexample to ensure users under 18 do not see regulated ads, such as ads for alcohol. As Snapisn’t capable of determining that a user is under 13 from our inferred age model, we have never applied it for the purpose of preventing users from registering an account. In line with our COPPA obligations, if Snap receives actual knowledge that a user is under 13 (for example, from averified parent’s request), Snap promptly deletes the account and associated data. Oversight We have a dedicated working group overseeing our age-assurance efforts. This cross-functionalgroup consists of 10+ FTEs with representatives from the product, privacy legal, product legal,policy, and trust and safety teams. These team members bring to the table extensive experienceand knowledge in the areas of operations, policy, global privacy laws and regulations,privacy-focused product decision making and online safety. Different iterations of this group meet in working groups that are focused on exploring,discussing and assessing possible ideas and concepts and solutions related to age assurancethrough risk assessments, discussions with our product team, and collaboration with externallegislators, regulators, vendors, experts, NGOs and other stakeholders. For example: ● This group regularly meets to discuss Snap’s age and parental assurance methods, andthe legal and regulatory requirements in this space. ● This team has assessed current industry practice with regards to age and parentalassurance, including in particular the mechanisms of: (i) device operating systems Google Family Link, Apple device parental controls and Family Sharing controls and Microsoft Family Safety and (ii) other online services such as Whatsapp, Tiktok, Instagram and Youtube. 263 CONFIDENTIAL ● The team regularly engages in stakeholder meetings, such as those organised by the Centre for Information Policy Leadership (CIPL) Privacy in Europe. ● This group also advocates for a holistic approach to age and parental assurance. Anexample of this would be our work with the UK Government and House of Lords duringthe passage of the Online Safety legislation in Parliament to successfully achieve arequirement for Ofcom (the UK’s communication regulator responsible for regulating theUK’s Online Safety Act) to consult holistically on child safety and age assurance, includingconsidering the role that infrastructure such as app stores / device operating systemaccounts have in providing privacy friendly, secure, effective and efficient solutions. ● This team has also met with a number of industry leading third party age assuranceproviders to assess the technical, legal, financial and user impacts of those services beingintegrated into Snap. ● This team has reviewed research from external stakeholders, on the positive and negativeimpacts of age and parental assurance, as well as consulting with users of Snapchat(including our own families and friends). ● Representatives from this group presented to Snap’s Safety Advisory Board77 on Snap’sapproach to age assurance and Family Center, including potential options and challengeswith age assurances. On age assurance in particular, this group of experts advised us that: ○ This was an industry wide issue that needed broad stakeholder discussion andneed for service independent solutions (i.e. considering devices, app stores etc)that best met the needs of children and parents/responsible adults. ○ They also felt our 13+ age limit was inhibiting our ability to recognise and keeppotential younger users safe and advised us to consider lowering the age soSnapchat’s U18 experience was available to those that wish to use and lessincentive for children and parents/responsible adults to try to access via 18+accounts / bypass age assurance methods. They recognised the challenges thatCOPPA presents in this regard. ○ They felt it was better to have a safe platform for all than to rely on excessive age /parental assurance. This would limit the need for age / parental assurance to thesmaller number of mature areas. Ongoing evaluation Snap continues to evaluate its approach, and consider possible concepts and approaches withindustry peers and third-party age-assurance vendors, to ensure we keep pace with industrypractice. We are also supporting legislators and NGOs in the UK, France, and elsewhere in the EUto enhance the role of app stores, online devices, and web browsers in providing appropriateinterfaces for age assurance and parental controls to facilitate consistent, effective and efficientapproaches for the online ecosystem. 77 The Safety Advisory Board is explained in Section 6.7 (Snap Advisory Groups). 264 CONFIDENTIAL Snap earned praise in recent years by the Children’s Advertising Review Unit (CARU) forexceeding minimal protections to keep underage users off Snapchat, and for providing numeroussafeguards for our users once they are on our platform. Despite this, we continue to reassess ourage assurance efforts, including engaging with industry partners, regulators, and third party ageassurance vendors, to ensure we keep pace with developments in the space. We activelyparticipate as thought leaders in industry roundtable forums, as well as with policy makers in theUS and Europe to understand the evolving landscape of age assurance, and its criticalimportance coupled with its inherent challenges. We’ve also held multiple exploratory anddeeper dive sessions with leading vendors across the age estimation and identity verificationmarketplace in recent months, as we consider third party technologies including biometrics, IDscan, and financial transaction methods (among others), to enhance our current approach. In terms of parameters, Snap consistently focuses on the potential impact to key areas whenevaluating the effectiveness of age assurance measures. Such factors include the tradeoffbetween safety of minors and compromising user privacy/data security, the accuracy andreliability of age estimation technology (particularly for younger or ethnic minority populations),the fairness of methods that may disadvantage users without official government IDs or bankaccounts, and the harm to industry competitiveness from the exorbitant cost of adopting thirdparty technology at scale. It is also worth noting that many organizations have concerns about the effects of introducing ageassurance and support device OS account and app store based solutions for age assurance tosupport online platforms’ own mitigation measures, for example ICMEC in the US. ProminentEuropean children’s NGOs have expressed similar support when meeting with Snapchat. We remain focused on thoughtful enhancements to our risk-based, age and parental assuranceapproach that include balancing the need for safety, accuracy, fairness, and user privacy amongother important factors and taking a holistic view of the online ecosystem used by children andtheir parent(s) (or other responsible adult(s)). However, currently, taking account the results of ourrisks assessment set out in Section 4 and current industry wide practices, we continue toconclude that our approach is proportionate, reasonable and effective. Commission Submission We have recently submitted our feedback to the Commission on its proposals for guidance onArticle 28 for online platforms. We support its goals and believe it is important we developeffective industry-wide standards for assessing a high level of privacy, safety, and security forTeens, in line with existing Age Appropriate Design Codes, Data Protection Impact Assessmentsand Privacy and Safety by Design obligations that already exist in Europe and other parts of theworld. We believe such a code should consider in particular the role of online platforms that are‘gateways’ (such as device operating systems, app stores and web browsers) through which 265 CONFIDENTIAL parents and Teens engage with such platforms. For example, it would make logical sense tofurther enhance age assurance at the App Store and device operating system levels to ensure arobust upstream solution78 and support the measures taken by third party services (likeSnapchat). Critically, Snap also consistently focuses on the potential impact to key areas when evaluating theeffectiveness of age-assurance measures. Such factors include the tradeoff between the safetyof minors and compromising user privacy/data security, the accuracy and reliability of ageestimation technology (particularly for younger or ethnic minority populations), the fairness ofmethods that may disadvantage vulnerable segments including users without official governmentIDs or bank accounts, and the harm to industry competitiveness from the exorbitant cost ofadopting third-party technology at scale. Transparency to Minors We continue to make efforts to provide users with information regarding our services in a waythat is clear and comprehensible across age groups. We verify readability of our key terms andconditions and privacy notices using automated readability tools. The vast majority of thesedocuments are shown to be understandable for our users. Our main terms and conditions is,necessarily, a formal legal document and contains longer provisions and more complex languagewhich our automated readability tools indicate may be more difficult for our younger users tounderstand. To improve readability in particular for our younger users, we arranged the terms andconditions into sections that provide a sensible flow, including appropriate and succinct sectionheaders, and added short summaries at the bottom of each section. We have tested thereadability of these short summaries, and confirmed they are understandable for our users,including Teens. In addition, we have provided short explainers for our Community Guidelines tofacilitate user understanding of this important document and know what they should and shouldnot be doing on Snapchat. Our Privacy Center was designed for our youngest Snapchatters and was intentionally developedto be easy to read and understood by all members of our community. We created our privacy andsafety hub, with pages such as our Privacy by Product page, to give Snapchatters a high-levelsummary of our privacy and safety practices across each of our products and features. We alsocreated a video to visualize our privacy practices, and use icons and other best practices asrecommended by privacy and safety experts and the recognised Age Appropriate Design Codes. As outlined in the Introduction, our European Snapchatter community consists of a diverse rangeof ages and genders. Snapchat services are not primarily directed at or used by minors. WhileSnapchat does have a young demographic, only a relatively small percentage of European Union 78 "Making Smartphones and App Stores Safe for Kids: Federal, State, and Industry Measures," published onNov. 16, 2023 by the Ethics and Public Policy Center and Institute for Family Studies: https://ifstudies.org/ifs-admin/resources/briefs/ifs-eppc-smartphonesappstoresbrief-nov23.pdf. 266 CONFIDENTIAL users fall within the 13-17 age category. The largest age category of European Union users is18-24. 5.8.3 App Store Level Safeguards App stores contain and facilitate a vast range of apps presenting a wide array of different risksand representing the entire spectrum of online risks. They have a special place in the ecosystemand therefore a uniquely high risk. They should thus be considered as ‘high-risk situations’. Inaddition to the specific risks presented by the individual apps hosted, app stores can generatehigher and exponential risks for a minor than the ones created by each individual service (e.g. aminor accessing harmful information on different services and combining harms). Larger app stores already recognise that they present unique risks and require additionalmitigations. In fact, larger app stores already apply age gates to prevent users from downloadingan app if their app store account age is below the app’s minimum category specified by the appprovider (and where applicable parents via the device operating system’s account level familycontrols - see below). Like most online platforms, app stores usually rely on the app storeaccount’s declared age (which is often the same account used by the device operating system).We also have noted that app stores rely on predetermined age categories which do notnecessarily capture nor fit the age categories defined by Snapchat and other developers at theapp level. A classical example is the app store category 12+, which does not align with the agethreshold of 13+ that is commonly specified by application services. There are improvements which could be made by app stores providers for the benefit of theentire ecosystem including: (i) stronger age assurance at the app store / device OS account leveland sharing that signal with developers to support their minor protection measures; and (ii)allowing developers to set a more precise minimum age. 5.8.4 Device-Level Safeguards Many additional controls are provided for teenagers, parents and other responsible adults. Manydevices now come with wellbeing settings, such as bedtime mode that turn off device and appnotifications and turn the screen the black and white to encourage sleep. 267 CONFIDENTIAL Additional controls are also provided via the device operating system’s account level family controls (e.g. Google Family Link, Apple device parental controls and Family Sharing controls and Microsoft Family Safety). For example, via these controls, parents and other responsible adults are able to view usage, set time limits, and disable access for each app which the teenagers haveon their device. 268 CONFIDENTIAL We have also noted that the providers that operate these family controls (who are alsogatekeepers pursuant to the EU Digital Markets Act) provide deeper levels of visibility and controlfor their own first party services. This level of interoperability and access would be very helpful forour own Family Centre (which is explained below) as it would increase the awareness andaccessibility for parents and other responsible adults who may not have a Snapchat account. Asexplained below, we are actively encouraging further multi-stakeholder dialogue to have drivesolutions that provide equal access and interoperability across the industry. 5.8.5 Platform-Level Safeguards There are several protections that we put in place at a platform level to mitigate the risk ofmalicious users of Snapchat. In particular, we have inference models in place that look at platform 269 CONFIDENTIAL wide meta-data signals to identify suspicious accounts. We use this information at a product levelto implement additional safeguards for Teen and adult users. 5.8.6 Product-Level Safeguards In addition to our defaults for all users, we have added protections in place for Teens, to helpmitigate risks in a number of ways. Public Content Once users decide to share a Snap via My Story, by default only friends can view it. Snapchatterscan choose to share to everyone, only to friends, or to a customized few. This emphasis onsharing with friends and giving users controls over who can view their content are in line withhow Snap takes into account privacy and safety when designing its features. Teen stories aredeleted by default. Their My Story view setting is defaulted to friends only. Friends lists areprivate. Viewing Public Content Teen accounts are restricted from access to certain content that is generally considered suitablefor 13+ but may contain certain shocking or sensitive content some may not find appropriate. Spotlight We have developed machine learning classifiers which work to identify sexually suggestivecontent and filter it from the experience before human intervention. In addition, our Spotlightcontent is evaluated by human moderators upon reaching a threshold number of views, andbefore being even more widely distributed. These steps reduce the likelihood of Teens accessingillegal or violating content, or content that may negatively affect their rights, security and health. We also aim to prevent older users from seeing content from younger users and to protect Teensfrom being contacted by older users. We seek to achieve this by, for example, implementing thefollowing measures: ● We limit the recommendation of content created by Teens to older users ● Adults cannot comment on Teen’s Spotlight content on Snapchat. ● Users can also choose to disable comments on any post. ● Teens are protected on Spotlight by not having their usernames displayed. Map As U18 users cannot have Public Profiles, they will not have their Public Stories featured on SnapMap when tagging a place or venue to a Public Story (which would occur for 18+ accounts). 270 CONFIDENTIAL Additionally, Map filters out suggestive content from being recommended to users ages 13-17 andage-gates certain types of locations to prevent them from showing on maps for minors, includingbars and tattoo parlors. Advertisements We restrict ads based on the user’s age. For example, ads for dating services must be targeted tousers over 18 and must not be provocative, overtly sexual in nature, or reference transactionalcompanionship. Similarly, ads for alcohol products must be age-targeted to at least 18+, or theapplicable minimum drinking age in the respective country where the ad is running. Advertisers must comply with our Ad Terms, Advertising Policies, and applicable national advertising codes. We prohibit ads that address or intend to appeal specifically to children under the age of 13. Reporting and Blocking ● In-App Reporting: Teens have the ability to report abuse they may observe or experiencewithin Snapchat. They can easily report Snaps, Chats, Stories, and Accounts by navigatingto the clearly marked “Report” option in the menu on each of these feature screens or bypressing and holding on the content itself. Users follow our simple reporting flow andprovide their reason for reporting and any additional comments that might be relevant.Reports are reviewed by our Trust \& Safety teams that operate 24 hours a day, 7 days aweek, and violating content and accounts are subject to enforcement. See theEnforcement section of this Report for more information. ● Blocking: All users have the option to Block another user. This prevents the friend from viewing friend content posted by, or sending Snaps and Chats to, the blockingSnapchatter. Since our 2023 Report, in an effort to prevent bullying and potential repeatharassment, we have introduced improvements to our blocking tools: Blocking a user willalso now block new friend requests sent from other accounts created on the samedevice.79 ● Removing Friends: All users also have the option to remove a friend from their friends list. Once removed, the Snapchatter will no longer be able to view content accessible only byfriends and, by default, should not be able to Chat or Snap. Friending ● Similar protections apply to prevent Teens from searching for unknown adults. We preventdelivery of a friend request altogether when Teens send or receive a friend request from 79 https://values.snap.com/news/new-features-to-help-protect-our-community 271 CONFIDENTIAL someone they don't have mutual friends with, and that person also has a history ofaccessing Snapchat in locations often associated with scamming activity.80 ● By default: Users need to accept bi-directional friend requests or already have each otherin their contact book to start communicating with each other. This design decision addsfriction and prevents users from communicating with each other prior to accepting afriend request or being in one’s contact book. ● No Public Friends Lists: Once users have accepted friend requests, the friend lists remainprivate. Snapchat does not disclose the friend lists of users to other users, nor do weexpose the total number of friends that a user has. This protects the privacy of the userand their friends. On most other platforms friend lists are public by default or there is anoption to share them publicly. These types of features create the ability for strangers tocontact vulnerable groups (e.g. younger users). ● Friend Check-Up: Prompts Snapchatters to review their friend lists and remove those theyare no longer in contact with, keeping their network up-to-date and focused on closefriends. ● In-App Warning: We provide pop-up warnings: (1) when a teen receives a message fromsomeone they don’t already share mutual friends with or have in their contacts81 and (2) ifthey receive a chat from someone who has been blocked or reported by others, or is froma region where the teen’s network isn't typically located.82 82 https://values.snap.com/news/new-features-to-help-protect-our-community. 81 https://values.snap.com/news/new-safeguards-for-snapchatters-2023. 80 https://values.snap.com/news/new-features-to-help-protect-our-community. 272 CONFIDENTIAL Family Center / Parent Tools Our in-app parental supervision tool, Family Center, gives parents, caregivers, and other trustedadults visibility into their teens’ friends list and who they have messaged with in the last sevendays, as well as the the ability to: (i) restrict their teen’s access to Spotlight and Discover contenttagged as 'sensitive' by our moderation team, (ii) limit their teen’s ability to engage with the My AIchatbot; and (iii) quickly request their teen’s location (which the teen must approve beforelocation is shared). Parents are also able to easily report accounts that may be in violation of ourCommunity Guidelines and have access to helpful resources directly in the app. Our goal in designing Family Center was to empower both caregivers and teens, balancingparents’ desire for more insight with teens’ desire for autonomy and privacy - notably ensuringthat teens’ messages remain private. We continue to put care and time into establishing thisbalance in a thoughtful way, engaging in user research and surveys, competitive research, focusgroups and interviews with both teens and parents, feedback sessions with dozens of onlinesafety experts and academics, including members of our current Safety Advisory Board, andextensive cross-functional internal reviews, including by our Product Legal and PrivacyEngineering teams. In their annual report,83 Jugendschutz.net, the joint competence center of the German Federaland State governments for the protection of children and young people on the Internet,highlighted Family Center as a positive example in the area of parental tools and support onsocial media platforms. The report concluded that Family Center can help teens and parents talkabout negative experiences, contacts, or time spent on the platform. At the same time, it notedthe opportunities for teen control, such as having to agree to parental guidance. 83 Jugendschutz, ‘Jugendschutz im Internet - 2022 Bericht’, April 2023, url. 273 CONFIDENTIAL 274 CONFIDENTIAL 275 CONFIDENTIAL 5.8.7 Conclusion We take the protection of Teens seriously on Snapchat. We’ve designed Snapchat to protect theirprivacy, safety, and security. Our key tenets include acting in the best interests of Teens, offeringstrict default settings for all users, and respecting Teens’ freedom to express themselves safely,while recognizing their right to information about the world. We aim to achieve these tenets bypositioning parents and guardians to help guide teens in their responsible use of our platform,attaching a heightened safety interest to Teens using our products, and establishing processes toensure we develop products in a way that upholds these tenets. We have implemented thesetenets through the use of Family Center, focusing on age-appropriate content, reporting and 276 CONFIDENTIAL blocking mechanisms, and putting in place appropriate protections and limitations on privatemessaging, friending, public content, and advertising. As explained in Section 4, we have concluded that the targeted measures we’ve taken toprotect the rights of the child, including age verification and parental control tools, incombination with the other mitigations explained in this Section 5, are reasonable,proportionate, and effective mitigation measures for the risks identified for Snapchat’sin-scope services. We continue to actively work with the Commission and others to establishan EU Age Appropriate Design Code and/or guidance on high level of privacy, safety andsecurity under Art 28 of the DSA, and consider whether further mitigation measures may bereasonable, proportionate, and effective for online platforms, ‘gateways,’ and other onlineservices. 5.9 Content Authenticity 5.9.1 Introduction Snap is aware that there is intense interest and concern surrounding the ways in whichadvancements in generative AI technologies are impacting online platforms. Snap recognises thepotential for AI generated or transformed content to be distributed through the inscope services of Snapchat, such as Spotlight and Discover, as explicitly called out in our Community Guidelines. Just like any other content distributed through these channels, this content may constitute illegalcontent or information that otherwise violates Snap terms and could contribute to the systematicrisks outlined in Section 34 of the DSA. 5.9.2 Risk Assessment Results Snap gave due consideration to the risks and harms that could arise from dissemination of usercontent in its risk assessment results section of this Report. In particular: ● Section 4.1.10 (Dissemination of harmful false information) - In this section, Snap recognised that “fake news,” (online) “disinformation” and “deep fakes” had gained a lotof attention in the media and academic and political debate over the last years. Werecognised that such content presented a risk of significant harm. This applied to allcontent formats, whether or not generated using AI tools. However, when consideringevidence relating to Snapchat specifically, we found very low prevalence rates of this typeof harmful content. We concluded that ‘harmful false information’ fell within the lowestlikelihood and risk prioritisation category relative to other harms being monitored onSnapchat. ● Section 4.3.1 (Negative Effect on Democratic and Electoral Processes) - In this section, Snap recognised that online platforms may have a negative effect on the electoral 277 CONFIDENTIAL processes and the exercise of political rights by amplifying digital disinformation ordeceptive content relating to political matters or processes. Again, this applied whether ornot generated using AI tools. However, when considering evidence relating to Snapchatspecifically, we found only limited occurrence of content harmful to democracy.Independent reports of electoral interference on Snapchat are vanishingly rare. Inconnection with a major, high-profile election in 2022, we onboarded Snap to the ElectionIntegrity Partnership (EIP),84 a partnership among leading research centers and civilsociety organizations who monitor online harms to democratic processes; as participantsin the EIP threat escalation program, our teams received only one single incident reportfrom the researchers monitoring risks on Snapchat. We concluded that ‘negative effect ondemocratic and electoral processes’ fell within our lowest likelihood and risk prioritisationcategory relative to other harms being monitored on Snapchat. ● Section 4.3.2 (Negative effect on civil discourse) - In this section, Snap recognised that digital content platforms could contribute to negative effects on civil discourse. Forexample, we noted: ○ The potential for personalized content and algorithmic biases lock users into echochambers, reinforcing existing beliefs and potentially leading to polarizedcommunities, which hinders open dialogue. ○ The risk of amplified dis- and misinformation negatively impacting public opinionon important civic issues. ○ The possibility of amplification of extreme or sensational content to retain userattention leading to heightened polarization and a hostile online environment. However, when considering the evidence relating to Snap specifically, we again found avery low prevalence of content related to harming civic discourse relative to othercategories being monitored. We concluded that ‘negative effect on civil discourse’ fellwithin the lowest likelihood and risk prioritisation category. 5.9.3 Mitigations Although there is not a high prevalence of harmful false information, fraud and spam orimpersonation, we take harmful information of this nature on Snapchat very seriously and Snaphas implemented a number of mitigation measures. Guidelines, policies, and practices Snap maintains robust policies––applicable to both the dissemination and the creation ofgenerative AI content––that function to mitigate risk and advance safety. 84 Election Integrity Partnership (2020), url. 278 CONFIDENTIAL Creation The features for creating generative AI content that Snap offers are not part of Snapchat’sinscope services and are out of scope of this Report (save for certain commonplace ad creationtools). Nevertheless, outside of its DSA obligations, we note that Snap has developed severalinternal policies relating to generative AI. In particular, (1) Content and Product policies: We have developed a suite of policies that disallow thegeneration of harmful content (including deceptive political content). Our policy andmoderation teams work in partnership with engineering and data science colleagues toensure that our AI products are responsibly trained on these policy parameters. (2) Acceptable Use: We have similarly developed Acceptable Use Policies that prohibit theuse of our AI tools to attempt to generate violative content at the prompt-level. With regards to the Content and Product policies, we have drafted and implemented internalGenerative AI Policies to govern the internal development of generative AI features, such asMyAI. Our Product and engineering teams refer to this policy as they train models or adaptmodels from third parties. Our Safety team applies this policy when testing new features. These aligned very closely with the rules for content dissemination, which are explained below. In addition, our generative AI tools feature a broad range of mitigation measures, depending onthe tool, and include for example: specific transparency statements, abusive language detectionand query related measures, age appropriate and/or canned responses, reporting mechanisms,off-by default settings, data minimisation, data sharing, testing and parental controls (see Section5.8 on the Protection of Minors). Although out of scope of the DSA and this Report, risks andmitigations relating to our generative AI tools are assessed via our privacy and safety by designproduct reviews (see Section 6.3). Dissemination In the context of dissemination of content on Snapchat’s online platform, in scope of the DSA, weunderstand well that online platforms may have a negative effect on the electoral processes andthe exercise of political rights by amplifying digital disinformation or deceptive content relating topolitical matters or processes. Our Community Guidelines and Terms of Service set out the rules on what content is allowed on Snapchat. They are focused on preventing harm to Snapchatters and the broader communityfrom content and behaviour, whether or not caused by generative AI or any other form of IT tools(such as Photoshop). These rules apply to all content formats across our platform, includingcontent that is AI-generated. While the rules are agnostic to content format or creative tools, theCommunity Guidelines specifically note: “We implement safeguards designed to help keepgenerative AI content in line with our Community Guidelines, and we expect Snapchatters to useAI responsibly. We reserve the right to take appropriate enforcement action against accounts that 279 CONFIDENTIAL use AI to violate our Community Guidelines, up to and including the possible termination of anaccount.” Our rules and internal enforcement guidance include clear provisions related to content risks, forexample for civic discourse and electoral processes. In particular, our Community Guidelinesprohibit spreading false information that causes harm or is malicious, such as denying theexistence of tragic events, unsubstantiated medical claims, undermining the integrity of civicprocesses, or manipulating content for false or misleading purposes (whether through generativeAI or through deceptive editing). Our Community Guidelines rules on false information refer to a more detailed Explainer that prohibits content that undermines the integrity of civic processes, or deep fake content or othermedia that is manipulated for false or misleading purposes. The Community Guidelines furtherexplain that these prohibitions extend to the following types of harmful content: ● Procedural interference: misinformation related to actual election or civic procedures,such as misrepresenting important dates and times or eligibility requirements forparticipation. ● Participation interference: content that includes intimidation to personal safety or spreadsrumours to deter participation in the electoral or civic process. ● Fraudulent or unlawful participation: content that encourages people to misrepresentthemselves to participate in the civic process or to illegally cast or destroy ballots. ● Delegitimization of civic processes: content aiming to delegitimize democratic institutionson the basis of false or misleading claims about election results, for example. Sharing such content will violate Snap’s Community Guidelines irrespective of whether it isAI-generated or user-generated, or whether it is generated on Snapchat or on another platform. Snap has a suite of internal policies and guidelines to help our content review and trust andsafety teams apply the Community Guidelines to user generated content disseminated via ouronline platforms (such as Spotlight and Discover). They provide more granular information for ourcontent review teams. For example, we explain that obvious jokes, memes, satire andnon-libelous comments about prominent social figures are OK; whereas false political narrativesmeant to undermine elections, or harmful / defamatory deepfakes, are NOT OK. In addition, our platform does not widely distribute an unvetted feed of algorithmically curatedpolitical information; we disallow all political content85 from Spotlight (our broadcast platform forUser Generated Content) unless it’s from trusted news partners and creators, and pre-moderate 85 For these purposes, “political content” means content related to political campaigns and elections,government activities, and/or viewpoints on issues of ongoing debate or controversy. This includes contentabout candidates or parties for public office, ballot measures or referendums, and political actioncommittees, as well as personal perspectives on candidate positions, government agencies/departmentsor the government as a whole. 280 CONFIDENTIAL that surface to ensure that other such political content is not distributed. This safeguard ensuresthat Snap is not algorithmically promoting political statements from unvetted sources, andgenerally reflects Spotlight’s function as an entertainment platform. (Consistent with ourcommitments to fundamental rights of expression and access to information, Snapchat providesother, non-algorithmically amplified spaces for users to express their views and politicalobservations, such as Chat and My Story; users can also seek access to political information fromknown publishers and creators whom Snap has on-boarded for distribution on the Stories tab). With regards to advertising, we do not require ads to label when advertisement includesgenerative AI content nor do require advertisers to disclose to us the tools they used to edit orcreate their ad creative. Instead, our approach is to subject all of our ads to a review process, andpolitical ads are also subject to fact checking. Deceptive ads are rejected, irrespective of whetherthey use AI, photoshop, or other digital editing tools. Ads that are not deceptive, and otherwisecomply with our Ad Policies, are approved to run (and if they are a political ad, they must includea “paid for by” disclaimer and are catalogued in Snap’s political ads library). User Guidance On-platform features for creating generative AI content are not part of Snap’s inscope servicesand are out of scope of this Report (save for certain commonplace ad creation tools). Nevertheless, outside of its DSA obligations, Snap has released a generative AI support site that explains what generative AI is and provides additional transparency around Snap’s practices withregard to generated images which are detailed below. Snap considers that there limitations and factors that need to be considered when addingwatermarks, labels and other marks to identify generative AI content: ● Noise: On Snapchat, AI is often on a spectrum of assistance rather than being the primaryfeature. If an identifying mark must be added any time any feature that uses any tool thatmight be classified as generative AI, even if in a minor way, that would be overwhelmingand users would likely start ignoring the watermark. Snap has adopted a focusedapproach. ● Fakery: whether visual or embedded in metadata, watermarks, labels and other markscan be counterfeited, altered or removed with a little technical skill. Snap thereforeconsiders the potential for activity to create undue credulity for fake content, or to erodetrust, and therefore avoids placing too great a reliance on watermarks, labels and othermarks. ● Privacy: C2PA-style metadata watermarking could risk compromising a user’s identity,location or other sensitive information. Snap takes this into account when consideringwhether to apply watermarks. 281 CONFIDENTIAL Given these limitations and factors, Snap often relies on other more proportionate, reasonableand effective measures to mitigate the risks presented by generative AI (such as contentmoderation). Nevertheless, Snap has chosen in some cases to indicate that a feature in Snapchatis powered by generative AI in a number of ways, including using the sparkle icon ✨, specificdisclaimers, Context Cards, or tool tips. When users see these contextual symbols or otherindicators in Snapchat, they should know that they are interacting with AI and not a human, orviewing content that has been produced using AI and does not depict real world scenarios. Here are examples of what these AI indicators look like in Snapchat: 282 CONFIDENTIAL ● When some AI generated images are shared with others on Snapchat, we may include aContext Card to let you know that the image was created with a generative AI-poweredfeature. ● Some generative AI-powered features, like Dreams and AI Snaps, allow you to create oredit images. When you export or save a generated image to Camera Roll, a watermark ofa Snap Ghost with sparkles may be added to those images. The purpose of thesewatermarks is to provide transparency that the image was created with generative AI andis not real or based on real events, even if it is a realistic style. Not all AI generatedimages will include a Context Card or watermark. Images created with non-Snap productsmay not be labeled as AI generated. Here’s an example of what the watermark looks like: 283 CONFIDENTIAL Enforcement Snap enforces these Community Guidelines fairly and consistently, using internal policies andguidelines, and applies outcomes that are commensurate with the severity of risk. Accounts that we determine are used to perpetrate serious, high-severity harms will immediatelybe disabled. For other violations of our Community Guidelines, Snap generally applies athree-part enforcement process: ● Step one: the violating content is removed. ● Step two: the Snapchatter receives a notification, indicating that they have violated ourCommunity Guidelines, that their content has been removed, and that repeated violationswill result in additional enforcement actions, including their account being disabled. ● Step three: our team records a strike against the Snapchatter’s account. A strike creates a record of violations by a particular Snapchatter. Every strike is accompanied bya notice to the Snapchatter; if a Snapchatter accrues too many strikes over a defined period oftime, their account will be disabled. This strike system ensures that Snap applies its policies consistently, and in a way that provideswarning and education to users who violate our Community Guidelines. The primary goal of ourpolicies is to ensure that everyone can enjoy using Snapchat in ways that reflect our values andmission; we have developed this enforcement framework to help support that goal at scale. Partnerships Snap has closely followed the negotiations on the EU AI Act and plans to continue to activelyengage and assess collaboration opportunities on the upcoming AI Pact, as well as on thedrawing of the related codes of practice for providers of general-purpose AI models and thoseregarding the detection and labelling of artificially generated or manipulated content. More broadly, tackling risks stemming from generative AI requires (among others) broadindustry-wide technical solutions which have not been clearly identified so far. This is why Snap isactively engaging with its peers and industry experts in different fora to share best practices andadvance the technical debate. These partnerships, industry collaborations and efforts include: ● OpenAI Integration: My AI is powered by OpenAI’s ChatGPT, and Snap closely partnerswith OpenAI in relation to providing the My AI service, including sharing feedback onmoderation of content. ● Tech Coalition / Working Groups on Generative AI: Snap is a member of the TechCoalition’s Working Group on Generative AI Content, and a member of the GenAI BriefingSubgroup. The Working Group on Generative AI Content meets regularly to facilitatedialogue and information- and idea-sharing around mitigating content-level generative AI 284 CONFIDENTIAL risks. The GenAI Briefing Subgroup meets periodically to plan expert briefings for TechCoalition members on topics related to Generative AI risks; such briefings have includedrepresentatives from government, law enforcement, civil society, and the researchcommunity. ● Tech Accord to Combat Deceptive Use of AI in 2024 Elections: Snap was an initialsignatory to the Tech Accord to Combat Deceptive Use of AI in 2024 Elections. Thiscompact seeks to set expectations for how signatories will manage the risks arising fromdeceptive AI election content created through their publicly accessible, large-scaleplatforms or open foundational models, or distributed on their large-scale social orpublishing platforms in line with their own policies and practices as relevant to thecommitments in the accord. The Accord was announced at the Munich SecurityConference in February 2024. ● ITI AI Futures Initiative: Through its membership in the Information Technology Industry Council (ITI), Snap has participated alongside other private sector actors in the AI Futures Initiative. Led by technical and policy experts spanning the tech ecosystem, the Initiative is a forum through which participants are developing action-oriented recommendationsfor AI policy and working to address emerging questions around AI. Deliverables to date have included the issuance of Global AI Policy Recommendations to help guide governments around the world as to develop responsible regulatory approaches toAI-related issues. ● HackerOne - Red-Teaming Collaboration: Snap partnered with HackerOne on redteaming exercises to test the strict safeguards Snap has in place around AI. Together withHackerOne, we made significant developments in the methodology for AI safety redteaming that has led to a more effective approach to surfacing previously unknownproblems. We refer to the HackerOne blog for more details: https://www.hackerone.com/ai/safety-vs-security ● As an active member of the EU Internet Forum, Snap will support the upcoming dedicatedworking group on generative AI matters. ● We are also members of the Centre for Information Policy Leadership (CIPL) and theFuture of Privacy Forum (FPF) which work with industry stakeholders (like Snap), NGOsand government agencies in each region to advance a broad array of information topics.CIPL has been a leader in AI matters for many years through its dedicated AI Project andspecific Brazilian AI Project. Most recently, in Europe, CIPL has responded to the UKInformation Commissioner’s Office (ICO)’s consultations on Generative AI, and led variousforums on Accountable Governance of AI and AI Regulation in Brussels and the UK.Similarly, FPF is working on AI Governance and other responsible Gen AI initiatives. 5.9.4 Conclusion Content authenticity is a very challenging topic without a silver bullet. Snap is very conscious ofthe issues and has implemented a number of measures (including relevant measures identified inthe Commission’s guidelines concerning elections). Snap continues to carefully monitor 285 CONFIDENTIAL developments and industry practice, including regarding whether and how best to use prominentmarkings and other measures to distinguish content that falsely appears to be authentic ortruthful. We have concluded that our position on the use of marks to distinguish content that falselyappears authentic or truthful is, in combination with the other mitigations explained in thisSection 5, reasonable, proportionate and effective for the risks identified for Snapchat’sin-scope services. 5.10 Trusted Flaggers 5.10.1 Trusted Flagger Program Snap’s Trusted Flagger Program was developed to help non-profits, non-governmentalorganizations (NGOs), select government agencies, and safety partners support the Snapchat community by leveraging a special channel to report content that violates Snapchat’s Community Guidelines. Trusted Flaggers send a completed report form with details of the potential violation via email to a dedicated, confidential email address. The email used is a high priority channel andreports are reviewed in less than 48 hours (with reports relating to the most serious harmsprioritized and reviewed well within this timeframe). Once a decision has been reached, Snapinforms the Trusted Flagger that appropriate action has been taken. This channel is reserved forurgently harmful situations and is designed to supplement in-app reporting, which is still verymuch encouraged. Our Trusted Flagger Program allows us to gain insight from the Trusted Flaggers over the typesof harm they are encountering, and the behavior of victims in these circumstances. In addition toproviding a specific reporting channel, the Trusted Flagger Program also allows us to build strongrelationships with Trusted Flaggers. Snap makes use of our strong relationship with TrustedFlaggers to give product safety updates, encourage the promotion of our safety tooling and provision of safety resources (like links to our Safety Center). Upon receiving notice from a DSC about a proposed Trusted Flagger, we will let the DSC know ifwe have any questions or concerns. Typically, when we are considering accepting a new TrustedFlagger we take into account geographic coverage, area of expertise, anticipated volume ofreports, among other factors. Once a Trusted Flagger is accepted we send them an onboardingpackage, which includes an overview document of the program, including our commitment toreview reports in less than 48 hours (with reports relating to the most serious harms prioritizedand reviewed well within this timeframe); instructions on how to file reports to our dedicated andconfidential email address; a metrics document to keep track of how many reports they’ve filedand among which harm categories, as well as contact information in case they have questions orconcerns about the program. 286 CONFIDENTIAL When a Trusted Flagger wants to file a report, they leverage our instructions on which categoriesof information they should include in their reporting email. After we receive an email, Snap’s Trustand Safety teams review the report and take any appropriate enforcement action, or requestadditional information if required for full investigation. Once a decision has been made, Snap willinform the Trusted Flagger that appropriate action has been taken as a result of the report. Our teams remain in contact with the Trusted Flaggers, including when we need to discuss anyissues with their reports. Our team evaluates the reports submitted based on the completion ofthe form, the accuracy of the information provided, and whether or not the report leads toenforcement or other action. If our team identifies trends that are impacting the quality of thereports that cannot be resolved with a Trusted Flagger, we will communicate this information withthe DSC to identify a resolution. We are monitoring the Commission’s publication pursuant to Article 22(5) of the entities that havebeen awarded the status of ‘trusted flagger’ pursuant to Article 22(2) DSA. 5.10.2 Conclusion Snap has an existing, carefully managed Trusted Flagger Program with valued memberorganizations from a wide array of countries including many in the European Union. Snap looksforward to evolving its Program to incorporate organizations that have been awarded trustedflagger status under the DSA. As explained in Section 4, we have concluded that Snap’s Trusted Flagger Programme, incombination with the other mitigations explained in this Section 5, is reasonable,proportionate and effective for the risks identified for Snapchat’s in-scope services. 5.11 Dispute Settlement Bodies 5.11.1 Overview and Approach We invest significant resources in our community support teams who work to resolve queries andcomplaints received from Snapchatters and others. In line with DSA requirements (Article 21), weare informing our users in Section 19 of our Terms of Service, entitled Dispute Resolution andArbitration, about the possibility of out-of-court dispute settlement in case they are not satisfiedwith the outcome of the internal complaint-handling system. As outlined above, we are alsoreferencing the possibility of out-of-court dispute settlement in our notices. As of 14 August, we are aware of four bodies that may have been established and certified by theDSCs. However, we note that the DSCs are required to notify the Commission that they havecertified these bodies and the Commission is required to publish the bodies on their websiteunder Article 21.8. However, as at 14 August 2024, we have not seen the Commission publish 287 CONFIDENTIAL such a list. As such we have not yet been able to confirm whether any bodies have been formallyestablished and certified by their DSCs. Once confirmed, users and others will have the option to contact such out-of-court disputesettlement bodies to raise their case. The relevant out-of-court dispute settlement bodies will then be able to reach out to Snap via our dedicated contact point (dsa-enquiries@snapchat.com) to start the out-of-court dispute settlement process. Snap will then engage, in good faith, with theselected certified out-of-court dispute settlement body with a view to resolving the disputefollowing Snap’s policies and procedures. 5.11.2 Enquires As of 14 August, we have received two queries from one body purporting to be a certified DSAout of court settlement body. 5.11.3 Conclusion Snap remains committed to resolving user disputes effectively and in line with DSA requirements.Furthermore, we continue to support establishing an EU-wide settlement body or an EU portal forbetter user interactions. This approach would ensure consistent application of rules across all EUmember states and provide a simplified, single point of access for operators. As explained in Section 4, we have concluded that Snap’s current approach to DisputeResolution, in combination with the other mitigations explained in this Section 5, isreasonable, proportionate and effective for the risks identified for Snapchat’s in-scopeservices. 5.12 Codes and Crisis Protocols 5.12.1 Cooperation Snap highly values cooperation with other providers and industry experts as a way to share bestpractices and learning experiences that can enhance risk mitigation strategies. We are highlycommitted to industry partnership to steer progress in the fight against illegal and harmfulcontent online. In particular, Snap is active member of the following groups: ● EU Internet Forum \- Snap is an active member and contributor of the EU Internet Forum (EUIF), which provides a collaborative environment for EU governments, the internetindustry, and other experts and partners to discuss and address the challenges posed bythe presence of malicious and illegal content online. The EUIF aims at exploring possibleresponses against abuse and exploitation of online platforms by terrorists and violentextremists, as well as other malicious actors, including those that groom children for thepurpose of sexual abuse and the production and dissemination of child sexual abuse 288 CONFIDENTIAL material online. Earlier this year, the scope of the EUIF work was expanded to tackle alsothe issues of drug sales online and the trafficking of human beings. ● Technology Coalition \- Snap is also a member of the Technology Coalition, which is an alliance of global tech companies who are working together to combat child sexualexploitation and abuse online. The Tech Coalition coordinates industry’s overall effort tocombat child sexual abuse online. It provides resources, education, and capacity-buildingto tech companies, and serves as a resource for external stakeholders - from globalpolicy-makers to members of the media - on what industry is doing to tackle this issue. ● WeProtect Global Alliance \- Snap is a Board member of the Weprotect Global Alliance, which brings together the private sector, government and civil society to drive positivechange to help protect children from sexual abuse online. ● Alliance to better protect minors online \- Until the recent discontinuation of the initiative in July, Snap was also a member of the Alliance to better protect minors online. This self-regulatory initiative was aimed at improving the online environment for children andyoung people by steering debates and exchanges on the topic. ● CIPL \- Snap is a member of The Centre for Information Policy Leadership (CIPL). This is aglobal privacy and data policy think and do tank based in Washington, DC, Brussels andLondon. We work with CIPL and other industry leaders, regulatory authorities and policymakers to develop global solutions and best practices for privacy and responsible use ofdata, including with respect to teenagers and young adults. ● The Future of Privacy Forum \- Snap is a member of The Future of Privacy Forum (FPF).FPF is a non-profit organization that serves as a catalyst for privacy leadership andscholarship, advancing principled data practices in support of emerging technologies. Wework with FPF on a range of matters, including developing best practices related toAugmented Reality (AR), Artificial Intelligence (AI), biometric data, children’s rights, andmore. ● Centre on Regulation in Europe (CERRE) \- Snap is a member of CERRE, which is anot-for-profit think tank based in Brussels. Its goal is to support and inform aboutregulation in Europe and beyond. We work with CERRE on in-depth reports and issuepapers that address the major regulation challenges and high-quality, policy-oriented research undertaken by top-level academics in the tech, media and telecom sector. Other Cooperation with industry Snap is actively involved in the work of a number of EU-based trade associations to contribute tothe policy debate to support the development of a proportionate regulatory framework topromote online safety. ○ DOT Europe \- Coordination on EU privacy, security,safety, content policy issues ○ ITI \- Coordination on EU privacy, security and safety policy issues 289 CONFIDENTIAL 5.12.2 Codes of Practice The DSA establishes that the Commission and the European Board for Digital Services (‘theBoard’) shall encourage and facilitate the drawing up of voluntary codes of conduct at Union levelto contribute to the proper application of the DSA (Article 45). Snap welcomes the opportunity to support industry-wide efforts to promote risk-mitigationpractices in the form of voluntary codes. As a company with limited resources, Snap is constantly required to prioritize and ensure itsresources and efforts are focused on where the biggest risks and challenges for the companyare. As we advance in our learning curve from our DSA risk assessment, we will continue toprioritize interventions where we see the highest risks. EU hate speech Code As part of its long-standing commitment to fight harmful and illegal content, Snap signed onto the EU Code of Conduct to counter illegal hate speech online in 2018. Since joining the code, Snap has successfully passed all the evaluations and in the course of the last monitoring exercise (2022), and for the 6th consecutive year, Snap did not receive any notification. Additionally, in the course of 2022 Snap has worked closely with the European Commission andother signatories to further strengthen some of the Code commitments by reinforcing and betterframing the existing cooperation between IT companies and CSOs, beyond the remit of the monitoring exercises. This work led to the publication of an Annex to the existing code in December 2022. Since February this year, Snap has been engaging with the European Commission team (DG Just)and regularly cooperating with other industry signatories with a view to contribute to the updateof the EU Hate Speech code to bring it in line with the DSA by 2024. FSM Code of Conduct In September 2017, Snap joined the ‘Freiwillige Selbstkontrolle Multimedia - Diensteanbieter e.V.’(FSM), an officially recognized voluntary self-regulation association for the protection of minors inonline media. EU disinformation code Snap has not yet signed up to be a member of the EU disinformation code. Being very resourceconstrained and considering our limited exposure to this type of risk, we have so far opted not tojoin. 290 CONFIDENTIAL Article 28 Guidance / EU AAD Code Since our VLOP designation, we engaged with different stakeholders at the EuropeanCommission86 to flag our interest to contribute directly to the discussion and working group onthe AAD Code. We understand focus has shifted to preparing guidance on Article 28 of the DSA for onlineplatforms on ensuring a high level of privacy, safety and security for minors. We have recentlysubmitted our feedback to the Commission on its proposals for guidance on Article 28 for onlineplatforms.87 We support its goals and believe it is important we develop effective industry-widestandards for assessing a high level of privacy, safety, and security for Teens, in line with existingAge Appropriate Design Codes, Data Protection Impact Assessments and Privacy and Safety byDesign obligations that already exist in Europe and other parts of the world. We believe any Article 28 guidance or European wide AADC should consider in particular therole of online platforms that are ‘gateways’ (such as device operating systems, app stores andweb browsers) through which parents and Teens engage with such platforms. For example, itwould make logical sense to further enhance age assurance at the App Store and deviceoperating system levels to ensure a robust upstream solution and support the measures taken bythird party services (like Snapchat). Critically, Snap consistently focuses on the potential impact to key areas when evaluating theeffectiveness of age-assurance measures. Such factors include the tradeoff between the safetyof minors and compromising user privacy/data security, the accuracy and reliability of ageestimation technology (particularly for younger or ethnic minority populations), the fairness ofmethods that may disadvantage vulnerable segments including users without official governmentIDs or bank accounts, and the harm to industry competitiveness from the exorbitant cost ofadopting third-party technology at scale. We see the proposed Article 28 guidance, or any European wide AADC, as an excellentopportunity to develop an industry-wide solution on age assurance, which could effectivelysupport the DSA implementation (drawing support from DMA to ensure interoperability andportability is provided by gatekeepers). We believe that such a solution should be developed atdevice level by operating systems, which would create a signal and make it available to services 87 The feedback can be found here:https://ec.europa.eu/info/law/better-regulation/have-your-say/initiatives/14352-Protection-of-minors-guidelines/F3496394_en 86 Meeting with DSA DG Connect F2 team in February and June 2023; Meeting with DDG Renate Nikolay inMay 2023; Meeting with Cabinet Suica in May 2023; meeting with EVP Commissioner Vestager in June2023; exchanges with DG Connect team G3 in June and July. 291 CONFIDENTIAL operating in their environments. Our proposed approach has attracted the interest of differentregulators, including recently in the UK, France and Germany (amongst others), and we would behappy to further expand on that if helpful. We have also been proactive in driving industry widediscussion through initiatives such as the CIPL / WeProtect Multi-stakeholder Dialogue on AgeAssurance. 5.12.3 Crisis Protocols Snap has set up a number of crisis management protocols to help the organization swiftly tackleunexpected incidents and help minimize their impact on our service, users and operations. Cooperation with external stakeholders is a very important element of risk mitigation for Snap.Knowledge sharing and best practice development with experts and peers are key to strengthenand increase the effectiveness of our internal risk mitigation measures. This is why the companyhas signed up to several voluntary codes and is actively engaged in many different internationalfora and associations to steer constructive debate and best practice development in areas likeCSEAI, protection of Teens, and hate speech (and we work closely with our global trusted flaggernetwork on these matters). We will continue to monitor our risks and prioritize interventions onthe most severe risk areas. When it comes to dealing with unexpected events resulting inheightened levels of risks for the platform, our Content Crisis Response Protocol plays animportant role in providing a structure to our collaborative internal operations and efforts. As explained in Section 4, we have concluded that Snap’s current approach to codes ofpractice and crisis protocols, in combination with the other mitigations explained in thisSection 5, is reasonable, proportionate and effective for the risks identified for Snapchat’sin-scope services. 6\. Ongoing Risk Detection and Management Snap has developed a number of practices to detect and manage risks to Snapchat’s in-scopeservices. This includes: (1) the establishment of a Platform Risk Framework based on Snap’sproduct values, established international human rights principles, and risk-based metrics such asprevalence and severity analyses; (2) the designation of a senior, cross-functional teamresponsible for applying the framework and assessing its outcomes, including a DSA GovernanceGroup and meeting; (3) development of a repository of internal resources to support thedetection and management of risk––these include harm severity assessments; prevalencemetrics; reporting data reviews and a library of Terms and policy resources; and (4) continualimprovement and assessment through our Digital Well-Being Index (DWBI) Initiatives and SafetyAdvisory Groups (Safety Advisory Board and Council for Digital Well-Being). 292 CONFIDENTIAL This Section of the Report provides further details of these practices pursuant to Article 42.4.(b)(reporting on the mitigation measures relating to Article 35.1.(f)) and Article 42.4.(e) of the DigitalServices Act. 6.1 Platform Principles-based Framework Snap has implemented a platform risk framework that draws on a combination of Snap’s productvalues, established international human rights principles, and risk-based metrics such asprevalence and severity analyses. The framework is divided into two parts that borrow from relevant, longstanding elements of theinternational human rights framework: (1) identification of core platform governance values; and(2) a set of balancing principles for weighing those values against risks to our community andother harms. Reference to both of these elements in conjunction with one another provides aconsistent approach for responsibly reviewing proposed harm mitigations with attention tofoundational values. As a result, we have a responsible, rights-respecting approach to platform governance anddetecting and managing risk. The DSA Compliance Team and Cross-Functional Working Groupsreview Snap’s approach periodically to ensure it is in line with DSA requirements and global bestpractices. 6.2 DSA Compliance Team and Cross-Functional WorkingGroups This section sets out Snap’s approach to the establishment of governance mechanisms overSnap’s compliance with the Digital Services Act (DSA), fulfilling the requirements set out inArticles 11, 12, 13, 41, and 43. 6.2.1 Introduction This section outlines Snap’s governance mechanisms relevant to compliance with therequirements of Regulation (EU) 2022/2065 on a Single Market For Digital Services andamending Directive 2000/31/EC (the “Digital Services Act” or “DSA”), in particular with regard tothe requirements outlined in Articles 11, 12, 13, 41, and 43. 6.2.2 DSA Independent Compliance Function Snap has established an Independent Compliance Function (ICF) as part of the Snap Legal team,separate from Snap operations, with sufficient authority, stature, and resources that is responsiblefor coordinating, overseeing, and implementing Snap's Privacy and Regulatory Program. The ICFprovides oversight to ensure the necessary internal processes, resources, testing, 293 CONFIDENTIAL documentation, or supervision are in place for compliance with the DSA and monitors Snap'scompliance with the DSA, ensuring the identification and mitigation of risks associated with Snapoperations. 6.2.3 Independent Compliance Function Leads Snap has designated Independent DSA Compliance Officers who report to the ManagementBody and fulfill the tasks outlined in DSA Article 41 for the head of compliance. Snap's complianceofficers possess the professional qualifications, expertise, experience, and capabilities necessaryto fulfill the designated responsibilities. The Independent Compliance Officers are alsoresponsible for monitoring Snap's compliance with commitments outlined in the relevant codes ofconduct or crisis protocols. The compliance function significantly overlaps with the Data Compliance and Data ProtectionOfficer function, including the requirements to closely cooperate with the EC, responding toinquiries and monitoring compliance with the DSA, and conducting risk assessments. The head ofthe compliance function reports to individuals who are members of Snap’s exec meetings andfrequently updates the Board and executives on DSA related matters. This satisfies the Article 41requirement for the head of the DSA compliance function to report directly to the management ofthe company. Snap’s Board of Directors appointed Snap's compliance officers in their 7/24/23 meeting. 6.2.4 Compliance Officer Qualifications Snap’s Compliance Officers have numerous years of experience in data protection, privacy,compliance and governance. 6.2.5 Operation of the Independent Compliance Function Responsibilities of the Independent Compliance Function The Independent Compliance Function is responsible for the following activities: - Cooperating with the relevant Digital Services Coordinator and the Commission for thepurpose of DSA compliance; - Providing oversight over the development of the Systemic Risk Assessment methodologyand the conduct of the Systemic Risk Assessment, ensuring it is conducted on the basisof the best available information and scientific insights - Ensuring that risks referred to in DSA Article 34 are identified and properly reported onand that reasonable, proportionate and effective risk-mitigation measures are takenpursuant to Article 35; - Organizing and supervising the activities related to the independent audit pursuant toDSA Article 37; 294 CONFIDENTIAL - Providing oversight over the validation of controls leveraged to mitigate risks, evaluationof controls, and review of policies; - Monitoring the compliance with DSA obligations; - Reviewing and approving transparency reports; - Informing and advising the management and employees about relevant obligations underthe DSA; - Where applicable, monitoring the compliance with commitments made under the codes ofconduct pursuant to Articles 45 and 46 or the crisis protocols pursuant to Article 48 of theDSA. Oversight and Monitoring of Snap’s DSA Compliance The Independent Compliance Function utilizes multiple ways to exercise oversight andmonitoring over Snap’s DSA compliance. Any relevant issues or observations are discussed within the Independent Compliance Functionas well as with relevant stakeholders and escalated to the Management Body and the CEO asneeded. The Independent Compliance Function investigates root causes of issues orobservations, develops mitigation plans and works with stakeholders and control owners toimplement such corrective actions. The Independent Compliance Function may also escalatesignificant issues / observations to the Management Body and the heads of the respectivefunction if the root cause is identified as relating to that function. DSA Management Body Snap has designated a Management Body which oversees and supports the independence ofthe compliance function and manages issues as escalated to the Body. 6.2.6 DSA Cross-Functional Governance Team Snap established a Cross-Functional DSA Governance team inclusive of senior personnel, whichincludes Legal, Public Policy, Product, Engineering, Trust \& Safety, and Information Securityteams. Given the high stakes related to DSA compliance, given the multi-faceted nature of DSArequirements and given cross-functional ownership and responsibilities, it is important tointroduce a DSA governance structure that reflects the complexity of the DSA. The purpose of this cross-functional DSA Governance Team is to ensure the cross functionalteams activities continue to align with the requirements of the Digital Services Act. Based on theproblem areas and findings we have instituted numerous types of changes including productexperience and design changes, extension of detection and enforcement mechanisms, policyand operational process changes, introduction of support and educational resources for users 295 CONFIDENTIAL and opportunities for users to flag certain types of harmful content or seek redressalmechanisms. The DSA Governance Team is responsible for managing, overseeing, monitoring, assessing andadjusting Snap’s DSA compliance program. The DSA Governance Team meets on a monthly basis(the Team and/or its members might meet more frequently if necessary). 6.2.7 Points of Contact Designation, Publication, and Change Management Upon designation, Snap's cross-functional DSA governance team established a process fordesignating and updating in a timely manner the publication of points of contact for regulatoryauthorities, recipients of the service, and the legal representative in a location that is publiclyavailable and easily accessible. Snap’s Cross-functional DSA Governance Team reviews Points of Contact and LegalRepresentative designation. If a new Point of Contact or Legal Representation is appointed, theData Compliance Officer or designee will work with the web team to update the above websiteaccordingly. Points of Contact are made publicly available in all official languages of the EU and are easily accessible on: https://values.snap.com/privacy/transparency/european-union Point of Contact for the Authorities Upon designation, Snap's cross-functional DSA governance team designated a single point ofcontact email address for communication with Member State authorities, the Commission, andthe Board to enable direct and effective communication between Snap and the relevantauthorities on matters related to the Digital Services Act. Upon designation Snap's cross-functional DSA governance team ensured the easily accessiblepublication of the relevant information relating to the point of contact for regulatory authorities,including the languages to be used in such communications, on the Snap website. Authorities can contact Snap at dsa-enquiries@snapchat.com, through our Support Site, which supports all official languages of the EU, and at Snap B.V. Keizersgracht 165, 1016 DP Amsterdam, The Netherlands 296 CONFIDENTIAL The European Commission also received contact details to communicate directly with the Headof Snap’s DSA Compliance Function on all matters. Law Enforcement can contact through mechanisms described here: https://www.snapchat.com/lawenforcement Point of Contact for Users Upon designation Snap's Cross-functional DSA Governance team designated a point of contactfor users to contact Snap regarding DSA-related matters that is user friendly. Upon designation, Snap made information on the customer service POC publicly available in aplace easily accessible to the user in a place where they would be expected to be and up todate. For general DSA inquiries, Snap can be reached through the dsa-enquiries@snapchat.com email address as well as by submitting a ticket through our Support Site. The support site is available to support users in all official languages of the EU. Legal Representative Snap's Cross-Functional DSA Governance team has designated a legal representative (Snap B.V.)in one of the Member States where the service is provided for complying with DSA obligationsand enforcement matters. This designation includes allocation of resources and authoritiessufficient to cooperate with relevant authorities and comply with decisions issued by theEuropean Commission in relation to the DSA. We have notified our Digital Services Coordinator of the contact information for our legalrepresentative, including the name, mailing address, phone number, and email, and ensured thepublication of the information in a publicly available location. Snap has appointed Snap B.V. as its Legal Representative for purposes of the DSA. Snap’s legalRepresentation can be contacted at Snap B.V. Keizersgracht 165, 1016 DP Amsterdam, The Netherlands 6.2.8 DSA Supervisory Fee In its latest assessment, the European Commission has determined that Snap does not meet thethreshold for the supervisory fee, and Snap’s current contribution is set at EUR 0. 297 CONFIDENTIAL 6.3 Privacy and Safety by Design 6.3.1 DSA Risk Management As explained in Section 3, Snap has diligently identified, analyzed and assessed systematic risks relating to Snapchat’s in-scope services, and has specified the mitigations it has in place toaddress them, as required by Articles 34 and 35 of the DSA. Snap recognises that, as well as carrying out this assessment annually, it must also re-assessprior to deploying functionalities that are likely to have a critical impact on the risks identified (andtherefore the mitigations specified to prevent them). This is required by Article 34 of the DSA, butthis is also an industry standard practice to ongoing risk management and found in many otherlaws requiring risk assessments (including guidance relating to DPIAs). Snap has developed a number of practices to detect and manage risks to Snapchat’s in-scopeservices. As explained in this Section 6, these practices include: (1) the establishment of aPlatform Risk Framework based on Snap’s product values, established international human rightsprinciples, and risk-based metrics such as prevalence and severity analyses; (2) the designationof a senior, cross-functional team responsible for applying the framework and assessing itsoutcomes; (3) development of a repository of internal resources to support the detection andmanagement of risk––these include harm severity assessments; prevalence metrics; reportingdata reviews and a library of Terms and policy resources; and (4) continual improvement andassessment through our Digital Well-Being Index (DWBI) Initiatives, Safety Advisory Board, andCouncil for Digital Well-Being.88 These are covered in other parts of Section 6 of this Report. Additionally, Snap has a Privacy and Safety by Design review process. Privacy and safety bydesign is a cornerstone of Snap’s approach to designing and launching its products, and is builtinto Snap’s compliance program. Snap has an extensive privacy and safety by design reviewprocess to assess privacy and safety risks in the design and development of Snapchat. As part ofits privacy and safety by design program, Snap documents a review prior to new product andfeature releases that materially affect the privacy, safety and/or security of its users. The privacyand safety by design process is a collaborative and cross-functional process, and stakeholdersfrom Snap’s legal and privacy engineering teams are embedded in key phases of the product’sdevelopment. At Snap privacy and safety by design decisions are typically made by cross-functional teams. Wehave a long standing cross functional team across Product, Eng, DSA Compliance Officer,Operations, Policy, Legal, Comms, Trust \& Safety, and Privacy teams which meets very regularly toaddress risks flagged through various mechanisms such as industry reports, current events/news,internal data analyses and investigations, and feedback from regulators to assess problems, 88 https://values.snap.com/news/inaugural-council-digital-well-being. 298 CONFIDENTIAL prioritize them and agree on strategy and execution plans to resolve identified risks. Internally theteam is called the Safety XFN. Every quarter the team meets in person and virtually to align onpriorities for the next quarter, and reflects on safety improvements that were made in priorquarters. These findings are also presented to senior leadership on a regular cadence. 6.3.2 Privacy and Safety by Design review process Privacy and safety by design is a cornerstone of Snap’s approach to designing and launching itsproducts, and is built into Snap’s compliance program. As highlighted above, at Snap, our missionis to empower people to express themselves, live in the moment, learn about the world, and havefun together. We believe that privacy and safety are foundational to the success of our mission. Additional information can be found in Section 5.1. Snap already had an extensive privacy and safety by design review process to assess privacyand safety risks in the design and development of Snapchat prior to the DSA coming into force,and this continues to be the case. As part of its privacy and safety by design program, Snapdocuments a review prior to new product and feature releases that materially affect the privacy,safety and security of our users. 6.3.3 Holistic Digital Risk Management Risk assessments and mitigation obligations are being an increasingly common tool of digitalservice regulation. In Europe, such obligations are not only imposed by the DSA, but also, forexample, GDPR (in the form of Legitimate Interest Assessments (LIAs) and Data Protection ImpactAssessments (DPIAs)) and also in the UK and several EU Member States, the Age AppropriateDesign Code (AADC) Assessments (or their equivalent) and shortly the UK Online Safety Act’ssafety assessments. It is important that Snap is able to manage these, often overlappingEuropean requirements (in addition to other global requirements) in an efficient, effective andoperationalised manner. In order to meet Snap’s annual and ongoing holistic risk assessment obligations, including withrespect to the DSA, Snap continues to use a Digital \& Data Impact Assessment (“DDIA”)framework that combines our privacy, safety and security obligations into a single riskassessment. This was designed prior to the DSA coming into force, and will in future cover otheraspects such as the EU AI Act. The DDIA includes a template that serves as a vehicle to conduct the various risk assessments.This template supports the consideration of risk and mitigations related to a specific Snapchatproduct and includes guidance for the consideration of key factors and influencers on that risk,such as the performance of Snapchat recommender systems, the intentional manipulation of theplatform, and regional and linguistic considerations. 299 CONFIDENTIAL The DDIA templates are implemented at the product level and cover a range of requirementsbeyond the scope of the DSA. The DDIA is embedded with our existing privacy and safety by design process and requires ourcross-functional team to consider if a product change results in a significant impact on ourexisting consolidated DDIA assessment (including the DSA aspects). If so, this is required to bere-assessed before the product change launches. The DDIA is completed dynamically dependingon the nature of the Snapchat feature and impact of the change being assessed. If reviewers determine that the change does require an update to the DDIA, they will work withSnap’s Legal team (and other cross-functional Compliance team that they engage as needed) toupdate the relevant DDIA accordingly. 6.3.4 DSA Critical Impact Check Snap recognises that, as well as carrying out our annual DSA Risk Report, it must also re-assessrisk prior to deploying functionalities that are likely to have a critical impact on the risks identified(and therefore the mitigations specified to prevent them). This is required by Article 34 of theDSA, but this is also an industry standard practice to ongoing risk management and found inmany other laws requiring risk assessments (including guidance relating to DPIAs). As part of the DDIA update review, product reviewers consider whether the change amounts to acritical impact. As a result, we are able to detect and manage our DSA risk assessment and mitigationobligations on an ongoing basis. Since our 2023 Report, we have not identified any deployedfunctionalities that were likely to have a critical impact on our assessment of risks and mitigationspursuant to Articles 34 and 35 of the DSA. 6.4 Prevalence Testing A key measure we have in place to holistically detect and manage risk of illegal and other harmfulcontent on an ongoing basis is prevalence testing i.e. testing the ‘Policy Violating Prevalence’(PVP) of Stories accessible to the public via random sampling. The sampling allows us to estimatethe percent of policy-violating views and monitor the presence of illegal and other violatingcontent on Snapchat. Through this prevalence testing, we are able to uncover blindspots andprioritize efforts to close those gaps through improvements to our proactive detectionmechanisms, infrastructure improvements and agent training. 6.4.1 Overall Our overall PVP metrics demonstrate that the effectiveness of our proactive detectionmechanisms, agent training and other content moderation and enforcement efforts has increasedsignificantly since our 2023 Report. 300 CONFIDENTIAL 6.4.2 Example mitigations In our previous report, we identified the following examples of mitigations we have taken toreduce prevalence An example of the further mitigations we have taken since our 2023 report, is an improvement tothe efficacy and reduced latency of content enforcement mechanisms. This has involvedmeasures to temporarily reduce exposure of content suspected of being illegal or otherwiseviolating our terms while in review. The vast majority of the content to which these brieftemporary measures apply are subsequently enforced and as a result, this has helped us tofurther reduce the percentage of violating views of violating content. 6.4.3 Conclusion In conclusion, prevalence testing continues to be an extremely valuable measure for our ongoingdetection and management of content risks. Our ongoing efforts to improve our prevalencetesting and our mitigations to reduce prevalence of illegal and other violating content hasresulted in very significant decreases in PVP rates). As noted in the conclusion of our 2023Report, there is always more to do as we ultimately aim to reduce prevalence across all ourviolating content categories to as close to zero as possible. 6.5 External Request Monitoring and Review As noted in our 2023 Report, we produce a semiannual (every 6 months) Transparency Report, that captures our Community Guidelines enforcement data, law enforcement operations data,and copyright \& trademark data. The goal is to provide insight into our content moderation data,as well as our work with law enforcement and governments, in terms of how we work to keep ourusers safe. As we produce the report, we recognize shifts in our metrics (e.g., spikes ordecreases in content and account reports and enforcements) and utilize these to informheightened awareness from our moderation teams. Internally we also continue to reviewadditional breakdowns of this data and, in preparation for our DSA compliance, we continue toreview data relating specifically to the European Union’s individual Member States. We also continue to monitor advertising review rejections, advertising reporting andenforcements, ‘privacy, data protection and DSA’ requests and general community supportrequests. Since the DSA came into force for Snapchat on 25 August 2023, we have alsomonitored queries relating to compliance received via our dedicated dsa-enquiries@snapchat.com email address. This dedicated contact point is published on our website here, pursuant to Articles 11 and 12 of the DSA. As with our 2023 Report, we have continued to review and use this external request data tosupport the conclusions reached in this Report. 301 CONFIDENTIAL 6.6 Digital Well-Being Index (DWBI) Initiative 6.6.1 Introduction In the Spring of 2022, Snap launched a research project designed to gain insight into howGeneration Z teens and young adults are faring online. Our inaugural Digital Well-Being Index (DWBI), a measure of Generation Z’s online psychological well-being, was announced on Safer Internet Day 2023. The study asked about the risks and potential harms teens and young adultsare encountering online across all platforms, services and devices, not just Snapchat. Weconducted the research in six countries – Australia, France, Germany, India, the UK, and the U.S,which includes three of the largest European countries, two of which are in the EU) – and alsoincluded parents of teenagers between the ages of 13 and 19. Snap invests in this research to glean insights about the overall online risk landscape, and weseek to share those learnings with other key stakeholders across the ecosystem. Researchers,academics, safety-focused non-governmental organizations (NGOs), governments, lawenforcement authorities, parents, caregivers, and the general public, all stand to deriveknowledge and intelligence from these findings. In our 2023 Report, we explained that we had repeated and expanded this research in 2023 (“Year Two”). For more about Snap’s Digital Well-Being Index and research, see: Our website, as well as this explainer, the full research results, and each of the six country infographics: Australia, France, Germany, India, the United Kingdom and the United States. We took account of this and the previous year’s research when conducting our assessment of risk and mitigations ashighlighted in our 2023 Risk Report, and have continued to do so in this Report. In this section we provide highlights from the Year Three research and Generation Z’s views onreporting. In 2023, a total of 9,010 people participated in the study, and in 2024, respondentstotaled 9,007 across the six countries referred to above. Our Year Two deep-dive findings werehighlighted by several research institutions and organizations and, upon endorsement fromleading child-safety NGOs, we have repeated this portion of the study in 2024. 6.6.2 Snap’s Digital Well-Being Index - Year Three In this Report, we have continued our research during 2024 (“Year Three”). In addition torepeating our 2023 work, our Year Three industry wide research has also investigated teen’s andyoung adult’s attitudes and sentiments around reporting problematic content to platforms andservices, authorities and others. This offers insight into teens’ and young adults’ attitudes andsentiments around reporting problematic content to platforms and services, authorities andothers. 302 CONFIDENTIAL 6.7 Snap Advisory Groups 6.7.1 Introduction Snap launched a new Safety Advisory Board (SAB) in April 2022 with the aim of growing and expanding membership to include a diversity of geographies, safety-related disciplines and areasof expertise. In doing so, we initiated an application process, inviting experts and individuals fromaround the world to formally express their interest in providing guidance and direction to Snap on“all things safety.” The SAB Board was developed to educate, challenge, raise issues with, and advise Snap on howbest to keep the Snapchat community safe and counterbalance the online harms-dominatedexternal landscape. When appropriate, the SAB provides feedback on new products, features,policies, and initiatives before they are launched or released. The SAB and its individual membersdo not act as a representative or spokesperson for Snap, but rather as a collection ofindependent voices. The initiative helps to shape Snap’s approach to important safety issues andprovides Snap with strategic safety-related advice and guidance as Snap grows. Our Advisory Board currently stands at 19 members, based in 10 countries and representing 11different geographies and regions, 4 of the Members are EU-focussed. The board comprises 14professionals from traditional online safety-focused non-profits and related organizations, as wellas technologists, academics, researchers, and survivors of online harms. Members are experts incombating significant online safety risks, like child sexual exploitation and abuse and lethal drugs,and have broad experience across a range of safety-related disciplines. In addition, the Boardhas three members who are young adults and youth advocates. We selected these applicants toensure the Board has ready access to the all-important “youth voice” and viewpoint; to makecertain a portion of the Board includes committed Snapchat users; and to seek to balanceprofessional views with practical perspectives from a core demographic of the Snapchatcommunity. The SAB meets three times annually: twice virtually and once, in-person at Snapheadquarters for an in-depth strategy session to help prepare Snap for the coming year’splanning. 6.7.2 Updates Since our 2023 Report, we have continued to progress our work with the Snap Safety AdvisoryBoard and have also established a new Snap Council for Digital Well-Being. Safety Advisory Board In addition to the deep investment Snap is making in its SAB cohort, it periodically consults acadre of some 50 safety experts from around the world on new product features andfunctionality, policies, and initiatives. Snap also conducts periodic internal trainings and 303 CONFIDENTIAL learning-sessions, inviting external experts to help inform and educate Snap personnel working ina variety of safety disciplines about the overall risk landscape and Snap’s potential exposure. Thisyear, collaborations focused on the U.S.-based National Center for Missing and ExploitedChildren (NCMEC) on the topics of sextortion and improving CyberTip reports to NCMEC, as wellas smaller, executive-attended sessions with WeProtect Global Alliance, IWF, Thorn, and others.Snap will continue to invest in these and other external partnerships and relationships to helpbolster internal knowledge and awareness of the overall risk landscape. Snap Council for Digital Well-Being Snap formally launched a new youth-focused program in January 2024 - Snap's first “Council forDigital Well-Being (CDWB)”. The Council is a pilot program in the U.S. designed to encouragesafer online habits and practices among teens with the aim of having these young peoplechampion their knowledge and insights in their schools and communities. For the first interaction of this program, we opened applications from U.S.-based teens, aged 13 to 16. The inaugural cohort was selected and announced in May. Following two virtual monthly meetings, the group traveled to Snap HQ in Santa Monica, California, USA, in July 2024 for the first Council Summit. The 2-½-day program consisted of breakout sessions for both the teen cohort and their parents/chaperones, full-group discussions, and guest speakers. The teens alsogot a glimpse into working at a technology company, as they were treated to a 90-minute“speed-mentoring” session with 18 Snap employees representing different roles and teams. The Summit yielded interesting conversations and insights on topics such as online pitfalls,parental tools, and the differences and similarities between digital and in-person social dynamics.By the end of the Summit, the full group, chaperones included, was extremely motivated to bemore involved in their own local communities and to act as ambassadors for online safety. Weshared with them some of our outreach material to aid their efforts, including the belowinfographics on reporting and Snap Safety Milestones: 304 CONFIDENTIAL 305 CONFIDENTIAL The program will continue through next summer, and we look forward to continued engagementwith this cohort, their parents and chaperones, and the Snap SAB. 6.8 Audit We recently completed the external DSA audit of Snap’s compliance with its obligations underChapter 3 of the Digital Services Act for the audit period between August 25th 2023 and June30th 2024 pursuant to Article 37. It is also worth noting, as we did in our 2023 Report, that Snap already has another annual auditprocess in place. Snapchat is subject to an independent third party audit pursuant to Snap’s 2014FTC Consent Decree in the United States. This assesses Snap’s eight controls related to its datagovernance program. The preparation that goes into this audit is an ongoing endeavorthroughout the year, and ramps up further in the months leading up to the audit. This workinvolves stakeholders across the company, including Ops, Product, Engineering, CustomerSupport, Privacy Engineering, Legal etc, and provides another layer of external audit of Snap’spractices, and further reassurance that Snap’s practices are appropriate. 306 CONFIDENTIAL 7\. Conclusion This Report has been prepared to meet Snap’s obligation under Article 42(4) of the DSA and setsout the results of: (i) the risk assessment conducted by Snap pursuant to Article 34(1); and (ii) thereview of the specific mitigation measures that Snap has put in place to assess whether theymeet the requirements of Article 35(1) DSA. The risk assessment conducted by Snap identified, analyzed and assessed in accordance withArticle 34(1) DSA any systemic risks in the European Union stemming from the design, functioningor use of Snapchat’s in-scope services. Snap has also reviewed the specific mitigation measuresthat it has put in place to ensure they are “reasonable, proportionate and effective” for thespecific systemic risks identified by its risk assessment as required by Article 35(1). The results of the risk assessment and mitigation review are set out in Section 4 of this Report. The specific mitigation measures put in place by Snap are further detailed in Section 5 and Section 6 of this Report as required by Article 42(4). The Report shows that we have reasonable, proportionate and effective mitigation measuresin place and we continue to monitor a few areas to confirm that if additional measures arerequired Snap will act accordingly, as follows: 1. Dissemination of illegal or violating content: Since our 2023 Report, we have observeda further substantial reduction in the prevalence of content that is illegal or otherwise violating Snap’s Terms being disseminated on Snapchat’s online services in general. We have observed this content to now be at a very low level compared to the prevalence ofthis content on websites and other online spaces. For the first time, none of our illegal or other violating content categories were observed from our testing to have a prevalence rate that fell within our highest relative prevalence category. Within this very low level ofdissemination in general: a. We have continued to categorize three dissemination risk areas as falling withinLevel 1 risk prioritization for Snapchat’s in-scope services: (i) child sexual abusematerial, (ii) sale of drugs and (iii) credible imminent threats to human life, due torisk of severe harm each may cause. We have again confirmed we havereasonable, proportionate and effective mitigation measures for all three of thesecategories. As a result of these ongoing measures, all three are still assessed tofall within our extremely low likelihood category of the risks identified by Snap. b. We have continued to categorize five dissemination risk areas as falling withinLevel 2 risk prioritization for Snapchat’s in-scope services: (i) sale of weapons, (ii)terrorism, (iii) adult sexual crimes, (iv) harassment \& bullying and (v) glorification ofself-harm, due to the risk of serious harm each may cause. We have againconfirmed we have reasonable, proportionate and effective measures in place for 307 CONFIDENTIAL all five of these categories. As a result of these measures four are still assessed tofall within our extremely low likelihood category of the risks identified by Snap. Inthe case to terrorism content, we have noticed a slight increase in prevalencesince our last Report, and, while this is consistent with the October 7, 2023 Hamasterrorist attacks and ongoing conflict in Gaza and Israel, we continue to carefullymonitor this risk category to ensure the slight increase is temporary andprevalence remains very low. In the case of harassment \& bullying content,although we have observed a significant fall in prevalence and reporting, from ourtesting of the in-scope services of Snapchat, we will continue to carefully monitorthis risk category as we look to achieve further reductions. c. We have continued to categorize eight dissemination risk areas as falling withinLevel 3 risk prioritization for Snapchat’s in-scope services: (i) illegal hate speech,(ii) sale of other prohibited products or services; (iii) intellectual propertyinfringements, (iv) other adult sexual content, (v) violent or dangerous behavior, (vi)harmful false misinformation, (vii) fraud and spam and (viii) content relating to otherillegal activities. We have again confirmed we have reasonable, proportionate andeffective measures in place for all eight of these categories. As a result of thesemeasures six of these are still assessed to fall within our extremely low likelihoodcategory of the risks identified by Snap. With respect to fraud and spam andadult sexual content, since our 2023 Report we have observed a substantialreduction in prevalence rates and both now fall within a lower likelihoodcategory. We are continuing to monitor both categories as we look to achievefurther prevalence reductions (and to ensure the consistently higher proportionof ad rejections for fraud and spam are not an indication that further mitigatingmeasures might be required with regards to ads). 2. Negative effects on EU Fundamental Rights: We continue to categorize: (a) three risks tofundamental rights as falling within the Level 1 priority category for Snapchat’s in-scopeservices: (i) human dignity, (ii) data protection and (iii) child rights; (b) one risk as fallingwithin the Level 2 priority category: private life; and (c) three risks as falling within theLevel 3 priority category: (i) freedom of expression, (ii) right to non-discrimination andfreedom of religion and (iii) right to consumer protection. We have again confirmed wehave reasonable, proportionate and effective measures in place for all of thesecategories. We continue to actively participate in efforts to develop EU wide guidanceto assess if further industry measures are needed to address risks to child rights. 3. Negative effects on Public Security: We continue to categorize three risks to publicsecurity within the Level 3 priority category for Snapchat’s in-scope services: (i) negativeeffect on democratic and electoral processes; (ii) negative effect on civil discourse and (iii)negative effect on public security. We have again confirmed we have reasonable,proportionate and effective measures in place for all of these categories. 308 CONFIDENTIAL 4. Negative effects on Public Health: We continue to categorize: (a) two risks to publichealth within the Level 1 priority category for Snapchat’s in-scope services: (i) negativeeffect on children; and (ii) serious negative consequences on physical and mental wellbeing; (b) one risk within the Level 2 priority category for Snapchat’s in-scope services:negative effects on gender-based violence; and (c) one risk within the Level 3 prioritycategory for Snapchat’s in-scope services: negative effects on public health. We haveagain confirmed we have reasonable, proportionate and effective measures in place forall of these categories. As above, we continue to actively participate in efforts todevelop EU wide guidance to assess if further industry measures are needed toaddress risks to the protection of minors. It is Snap’s mission to reduce and maintain a lower prevalence of illegal and otherwise violatingcontent on Snapchat’s inscope services. We have been successful in increasing the granularity ofdata that we rely on, which we have incorporated into this Report. We will continue to progressthis action to ensure that risks can be tracked with even greater precision across in respect ofeach of Snapchat’s in-scope services. Regarding our Mitigations, since our 2023 Report, we have worked hard to: (i) evaluate our mitigation measures against new guidance from the Commission (in particular the Guideline forproviders of VLOPs and VLOSEs on the mitigation of systemic risks for electoral processes andthe updated EU Code of Conduct to counter illegal hate speech online) and (ii) provide additionalinformation on our measures to reflect the increased use of generative AI technology such asupdates to our content moderation policies (note that many of these measures relate to out ofscope services on Snapchat). Where required by the DSA, we have made available ourmitigations in all of the languages of the European Union and in all other cases in every languagein which Snapchat is available. Since our 2023 Report, we did not identify any deployed functionalities that were likely to have acritical impact on our assessment of risks and mitigations pursuant to Articles 34 and 35 of the DSA. As described in the Ongoing Risk Detection \& Management section above, our DSA Governance Team continues to regularly evaluate the effectiveness of its measures as we look tomaintain or further reduce prevalence, detect any new risks, assess any deployed functionalitiesfor critical impacts and determine whether further mitigating measures might be required. In summary, we have carried out our second annual risk assessment of Snapchat’s in-scopeservices required by Article 34(1) of the DSA. We have observed further significant reductionsin the prevalence across our illegal and otherwise violating content categories. We continue toconclude that we have reasonable, proportionate and effective mitigation measures, tailoredto the specific systemic risks identified, as required by Article 35(1) of the DSA. There arethree risk categories that we continue to monitor to ensure that remains the case. 309 CONFIDENTIAL 8\. Final Words This is the second year in which VLOPs have had to produce a report on their assessment ofrisks and the specific mitigation measures they have put in place. Snap has continued to take acomprehensive approach to the obligations in Articles 34, 35 and 42. As in our 2023 Report,although there is still no settled legal definition of ‘systemic risk’, we have again adopted theposition that all the risks identified in the DSA are systemic to online platforms (which is why theyhave been identified in the DSA). We are still then looking to ensure we have appropriateplatform wide measures in place in general, taking additional steps for specific risks, certainservices and high priority risks as necessary. As noted in our 2023 Report, our position reflects Snap’s own internal approach to riskmanagement and our core values to be kind, smart and creative. We have always taken theassessment of privacy and safety risks and mitigations seriously and this is demonstrated again inthis Report which concludes that Snapchat represents an even lower risk profile than identified inour 2023 Report. This is due to its unique design and function, but also in particular to the effortsof our cross-functional teams who have worked hard to provide high levels of privacy, safety andsecurity for all our users and further substantial falls in the risks specifically referred to in the DSA.We are particularly proud of the significant decrease in the overall prevalence rate for our illegaland harmful content categories. We were pleased to receive positive feedback from the Commission on our 2023 Report. Welook forward to again receiving feedback from the Commission on this second report, as well asthe publication of our first risk assessment and mitigation report and our audit report. 310 CONFIDENTIAL Annex 1 - Community Guidelines: Explainer Series Sexual Content Community Guidelines Explainer Series Harassment \& Bullying Community Guidelines Explainer Series Threats, Violence \& Harm Community Guidelines Explainer Series Harmful False or Deceptive Information Community Guidelines Explainer Series Illegal or Regulated Activities Community Guidelines Explainer Series Hateful Content, Terrorism, and Violent Extremism Community Guidelines Explainer Series Severe Harm Community Guidelines Explainer Series Snapchat Content Moderation, Enforcement, and Appeals Community Guidelines Explainer Series 311