row_id,org,type,founded,hq_country,what_it_does,lab_engagements,coefficient_total_usd,coefficient_grants_n,sff_total_usd,other_funders,lab_money,gov_contracts,leadership,named_in_sep_2026,source_urls,notes
EV01,METR (Model Evaluation and Threat Research),nonprofit,2022 (as ARC Evals inside ARC); spun out 2024-04-30,US,Runs pre-/post-release dangerous-capability and autonomy evaluations of frontier models and incident investigations.,"OpenAI (GPT-4 as ARC 2023; GPT-4o/o1/o3/GPT-5 series; Aug 2026 Hugging Face incident investigation, unpaid); Anthropic (Claude 2 onward; Feb/Aug 2026 Risk Report pilot external reviews; Sep 9 2026 eight-week incident investigation; named once in Amodei essay Sep 12); Google DeepMind (FSF v1.0 credit); Amazon (Nova Premier / Nova 2.0 Lite third-party assessor); Magic; G42 framework help. See evals.csv and lab_statements.csv.",0,0,4427000,"Audacious Project ~$17M of ~$38M Canary commitment (2024-10); Schmidt Sciences (undisclosed); Packard; Pew; Sijbrandij Foundation; Longview Philanthropy ($220,000 2023 + pooled); Effektiv Spenden (EUR 128,000 2023); Jaan Tallinn direct ($204,000 2024); Dylan Field; Steve Newman; Geoff Ralston; David Farhi; Jane Street individuals; LaCentra-Sumerlin; Astralis; Expa.org; FY2024 990 revenue $13,639,155 incl. $4,501,424 ARC transfer; ~$71M commitments raised Feb-Aug 2026",no (policy): metr.org/about says it does not accept donations from frontier AI companies or made by/at the direction of their employees; accepts free API credits (~$400K for HF investigation); HF investigation unpaid,"UK AISI (grant/contract, amount undisclosed); EU AI Office technical-assistance contract on loss-of-control (amount undisclosed, 'a small part of our income'); also consortium member of EquiStamp-led EU Lot 3 (EUR 1,167,484 total to consortium); no CAISI contract found","Beth Barnes, CEO (ex-OpenAI alignment, ex-DeepMind); Chris Painter, President (ex-US DoD JAIC)","Dwarkesh Patel Sep 9 ('ordain it the evaluator'); Anthropic Sep 9 (investigation agreement); Dario Amodei essay Sep 12 (example embedded evaluator); Joe Benton Sep 9/11 (joining METR); attacked by Sacks Sep 13, Khlaaf/Recht Sep 12",https://metr.org/about;https://x.com/dwarkesh_sp/status/2097707668920205580;https://darioamodei.com/post/we-must-pace-the-frontier;https://survivalandflourishing.fund/2023/h1/recommendations;https://survivalandflourishing.fund/2024/recommendations;https://survivalandflourishing.fund/2025/recommendations;https://gt990datalake-rawdata.s3.amazonaws.com/EfileData/XmlFiles/202523209349300367_public.xml;ted-864574-2025.xml,"Coefficient index: no grant to METR under any name ('metr' substring only hits Center for Welfare Metrics / IHME); ARC got $1,515,000 (2 grants, 2022) before the spin-out - counted on the ARC row. SFF = ARC Evals 2023-H1 $3,247,000 + METR 2024 $204,000 + METR 2025 $548,000 + $428,000 matching pledge. Local files used, nothing re-fetched."
EV02,Apollo Research,company (PBC since Jan 2026; previously fiscally sponsored nonprofit via Rethink Priorities; US 501(c)(3) Apollo Research AI Foundation EIN 99-4122618 has no filings),2023,UK (London; SF office 2026; DE-registered PBC),"Scheming/deception evaluations of frontier models, pre-deployment testing, plus a commercial agent-monitoring product (Watcher).","OpenAI: pre-deployment testing of o1 (2024), anti-scheming training study (2025); 'ran evaluations for all major labs' in 2025 per own timeline; policy engagement with EU AI Office, UK Frontier AI Taskforce. https://www.apolloresearch.ai/about",4410180,3,1133500,"Seed round (Jan 2026) led by 50Y with Juniper Ventures, Macroscopic Ventures, Common Metal, SAIF (Geoff Ralston's Safe AI Fund), Srikar Varadaraj, Sarah Meyohas, Ocean Investment, Progress Fund, Bryan Johnson (amount not stated); philanthropy continues for governance/research; SFF 2023-H2 via LSG $277,169 + via SFF $490,000 + via SFC $115,331; SFF 2024 $251,000 (via Rethink Priorities)",unknown: PBC post says it 'will continue to work with frontier AI companies to evaluate their models' and cites 'quickly growing market demand' for its evals; no published cap or refusal policy found,none with amounts found; policy recommendations to UK Frontier AI Taskforce / EU AI Office (advisory),"Marius Hobbhahn, CEO and co-founder (2023; TIME100 AI 2025); Charlotte Stix, Chief Governance Officer; Daniel Kokotajlo (AI Futures Project) first 'mission director' board seat",none found,https://www.apolloresearch.ai/about;https://www.apolloresearch.ai/blog/apollo-research-is-becoming-a-pbc;https://www.apolloresearch.ai/team;https://survivalandflourishing.fund/2023/h2/recommendations;https://survivalandflourishing.fund/2024/recommendations;https://projects.propublica.org/nonprofits/organizations/994122618,"Coefficient names matched: 'Apollo Research' ($1,535,480 2023 Startup Funding; $2,178,700 2024 General Support; $696,000 2026-05 Staff Support); 'Apollo Fellowship' ($321,838) excluded as unrelated. Team page shows EIN 93-4310599 (DE) which differs from the ProPublica foundation EIN; neither has a 990 on file."
EV03,Epoch AI,"nonprofit (501(c)(3) Epoch Artificial Intelligence Inc, EIN 99-4050541, exempt Sept 2024; earlier under Rethink Priorities)",2021 (volunteers); org 2022,US (San Francisco),"AI trends data, FrontierMath and other benchmarks, independent model evaluations and paid consultations.","OpenAI: commissioned FrontierMath (2024/25), FrontierMath Data Audit and Open Problems Verifier (2026); Google DeepMind: model evaluations joint report (2024/25), AI co-mathematician evaluation (2025); EU AI Office technical consultations (2026); ARIA (2025); UK DSIT; Stanford AI Index. https://epoch.ai/about/transparency",24513611,7,0,"Epoch transparency page (donations >= $70K): Coefficient (matches index); Survival & Flourishing Fund $195,000 (May 2026, not on any published SFF round page); Jaan Tallinn $600,000 DAF (Jan 2025); Leopold Aschenbrenner $200,000 via Manifund (Jan 2025); Likith Govindaiah $400,000 (Jan 2025); Girish Sastry $100,000 robotics (Dec 2025); Schmidt Sciences (FrontierMath Open Problems pilot, Dec 2025, amount not shown); Sentinel Bio $85,000 + $80,000; Carl Shulman $100,000 (2023); 2024 990-EZ revenue $180,010 (new entity)",yes: OpenAI paid for FrontierMath (disclosed Dec 2024/Jan 2025 after criticism; Epoch: 'we made a mistake' in late disclosure); Google DeepMind and OpenAI listed as paying consultation clients on transparency page; no cap stated,"EU AI Office technical consultations 2026 (amount not stated); consortium member of EquiStamp-led EU Lot 3 (EUR 1,167,484 to consortium); ARIA 2025; UK DSIT partnership","Jaime Sevilla, Director (Treasurer on 2024 990-EZ); Tamay Besiroglu listed as President on 2024 990-EZ",none found,https://epoch.ai/about;https://epoch.ai/about/transparency;https://epoch.ai/latest/openai-and-frontiermath;https://techcrunch.com/2025/01/19/ai-benchmarking-organization-criticized-for-waiting-to-disclose-funding-from-openai;https://projects.propublica.org/nonprofits/organizations/994050541,"Coefficient names matched: 'Epoch' (6 grants, $16,013,611) + 'Epoch AI' (1 grant, $8,500,000 2025-06). Epoch's own page dates/labels differ slightly from the index (e.g. $8.5M shown as April 2025; index 2025-06-06) and adds a $600,000 July 2026 compute grant not in the 2026-09-11 index snapshot."
EV04,EquiStamp,company (becoming a PBC),2023,"US (Lewes, Delaware)","Contract eval engineering and research operations for AI-safety labs (Inspect-based evals, baselines, annotation, red/blue teaming).","No direct lab evaluations found; builds infrastructure for evaluators: HCAST for METR, ControlArena for UK AISI, LinuxArena for Redwood (CEO bio). Lead tenderer of EU Lots 3 and 4. https://equistamp.com/",0,0,0,"private company; no funding disclosed on site; revenue from clients (METR, UK AISI, Redwood, EU AI Office)","unknown: no policy on site; clients named are evaluators/governments, not labs","EU AI Office TED 864574-2025: Lot 3 Loss of Control lead (EUR 1,167,484, with METR and Epoch), Lot 4 Harmful Manipulation lead (EUR 881,345, with Clarity AI Research Inc (Transluce) and CeSIA), awarded 2025-11-28 / 2025-12-04; subcontractor on FAR.AI-led Lot 1; UK AISI ControlArena work (amount not public)","Chris Canal, co-founder and CEO (ex-Amazon machine learning; also CEO of Measuring AI Progress)",none found (Canal quoted on standards bodies in July 2026 TechNewsWorld),https://equistamp.com/;ted-864574-2025.xml;https://www.far.ai/blog/far-ai-selected-to-lead-eu-ai-act-cbrn-risk-consortium;https://www.linkedin.com/in/chriscanal/;https://www.technewsworld.com/story/google-deepmind-ceo-calls-for-frontier-ai-standards-body-180439.html,Founder bio from a LinkedIn search snippet (equistamp.com/team returned 403). 'EquiStamp' appears nowhere in the Coefficient index or SFF pages.
EV05,Redwood Research,"nonprofit (501(c)(3) Redwood Research Group Inc, EIN 87-1702255)",2021,US (Berkeley),AI control and alignment research; staff subcontracted to METR for incident investigations.,Anthropic: alignment-faking paper 2024 with employee-level model access for Greenblatt; advises Google DeepMind (own home page); OpenAI Hugging Face incident investigation with METR Aug 2026 (unpaid); staff subcontracted to METR for the Anthropic investigation (Redwood X post Sep 12); UK AISI control safety-case work. See redwood.csv RW23-RW36.,63086000,5,2372000,"Jaan Tallinn direct $1,330,000 (2021-11 via FP-US); SFF 2022-H1 $1,274,000 and 2023-H1 $1,098,000 (only $598,000 paid per Tallinn ledger); 990 revenue $13.9M (2021), $12.0M (2022), $10.0M (2023), $22,060 (2024 after Constellation spin-out)",unknown: no published policy; Greenblatt received no payment from OpenAI for the HF investigation (RW32); Coefficient grants are the dominant income,UK AISI collaboration on AI-control safety cases (no amount); none with CAISI/EU found,"Buck Shlegeris, CEO and co-founder (ex-MIRI); Nate Thomas, co-founder and board chair; Ryan Greenblatt, Chief Scientist","none found; Sep 12 Redwood post disclosed staff subcontracted by METR for the Anthropic investigation; it did not propose Redwood as evaluator",redwood.csv;https://www.redwoodresearch.org/;https://coefficientgiving.org/grants/general-support-40/;https://survivalandflourishing.fund/2022/h1/recommendations;https://survivalandflourishing.fund/2023/h1/recommendations;https://projects.propublica.org/nonprofits/organizations/871702255,"From local redwood.csv; nothing re-fetched. Board overlap: Christiano, Karnofsky, Cotra all former Redwood directors (RW13-RW15). audit 2026-09-14: named_in_sep_2026 was a Redwood disclosure and Kokotajlo comment, neither of which proposed Redwood as evaluator."
EV06,Transluce,"nonprofit (501(c)(3) Clarity AI Research Inc, EIN 99-3262331)",2024 (launched Oct 2024),US (San Francisco),Open-source tools (Docent) and behavior reports for AI oversight; independent evaluations of frontier models.,"OpenAI: pre-release o3 truthfulness evaluation (Apr 2025, early access); GPT-5.1 Codex analysis; 77-model mental-health-crisis evaluation across OpenAI, Anthropic, Google, Meta, xAI, Thinking Machines, DeepSeek, Moonshot (Aug 2026). https://transluce.org/investigating-o3-truthfulness",0,0,0,"2024 990: contributions $6,870,000, net assets $5,214,033; donors not named on site; FY2025 revenue 6% from personal holdings of OpenAI employees and 32% from personal holdings of Anthropic employees (unrestricted, per its own policy page); board includes Mike McCormick (CEO, Halcyon Futures)","policy: 'may accept funding from AI developers or developer employees, including in exchange for evaluation services or products, but with limitations' - must disclose the share of revenue each year; no revenue 'directly from AI developers as organizations' in FY2025; no paid evaluations carried out to date; policy follows AI Evaluator Forum AEF-1 standard","EU AI Office Lot 4 (Harmful Manipulation) as consortium member under EquiStamp (EUR 881,345 to consortium), awarded 2025-12-04","Jacob Steinhardt, co-founder and CEO (UC Berkeley professor, on leave); Sarah Schwettmann, co-founder and Chief Scientist (ex-MIT CSAIL)",none found,https://transluce.org/about;https://transluce.org/independence-and-transparency-policy;https://transluce.org/donate;https://transluce.org/investigating-o3-truthfulness;https://projects.propublica.org/nonprofits/organizations/993262331;ted-864574-2025.xml,Not in Coefficient index under 'Transluce' or 'Clarity AI Research'. Transluce advisor Percy Liang (Stanford) and METR board member Rajiv Dattani (founding advisor) are cross-links.
EV07,Irregular (formerly Pattern Labs),company (VC-backed),2023 (as Pattern Labs),Israel / US,"Frontier AI cyber-capability evaluations (SOLVE / SOLVE+ scoring, FrontierCyber) and simulated network environments.","Anthropic: Claude 4 cyber evaluations as Pattern Labs (2025), SOLVE used for Claude 4 risk vetting; named as the third-party evaluation partner whose testing environment was involved in Anthropic's 2026 cyber-eval incidents; OpenAI: cited in o3/o4-mini and GPT-5 system cards; own GPT-6 Astra FrontierCyber assessment Sep 3 2026; UK government uses SOLVE. https://red.anthropic.com/2025/claude-4-cyber",6799999,1,0,$80M round Sep 2025 led by Sequoia Capital and Redpoint Ventures with Swish Ventures and Wiz CEO Assaf Rappaport; valuation ~$450M; Sequoia also co-led Anthropic's Series H,yes (inferred): commercial evaluation partner of Anthropic and OpenAI; 'Trusted by the world's leading AI labs' on its site; no independence policy published,UK government uses SOLVE framework (no contract amount found); none with CAISI/EU found,"Dan Lahav, co-founder and CEO; Omer Nevo, co-founder and CTO (prior affiliations not stated in fetched sources)",none found,https://www.irregular.com/about;https://techcrunch.com/2025/09/17/irregular-raises-80-million-to-secure-frontier-ai-models/;https://sequoiacap.com/article/partnering-with-irregular-ahead-of-the-curve/;https://www.anthropic.com/research/investigating-incidents-cybersecurity-evals;coefficient index 'Pattern Labs',"Coefficient name matched: 'Pattern Labs' $6,799,999 'Technological Risk Mitigation' 2024-02-28; no grant under 'Irregular'."
EV08,Nemesys Insights LLC,company,not stated (spin-out of University at Albany CART),"US (Albany, NY)",Red-teaming and strategic-analysis consultancy; CBRN threat assessment for AI model evaluations.,"Amazon: third-party assessor for Nova Premier (2025) and independent CBRN auditor for Nova 2.0 Lite (Jan 2026) under Amazon's Frontier Model Safety Framework (lab_statements L13, L16); EU Lot 1 CBRN subcontractor. https://arxiv.org/pdf/2601.19134",0,0,0,private LLC; no funding disclosed,yes (inferred): paid assessor for Amazon per Nova papers; no policy published,"EU AI Office Lot 1 subcontractor to FAR.AI consortium (EUR 1,434,080 to consortium; Nemesys share not public)","Gary Ackerman, CEO (director of UAlbany Center for Advanced Red Teaming); Brandon Behlendorf, Chief Science Officer",none found,https://www.nemesysinsights.com/ourstory;https://www.nemesysinsights.com/ourteam;https://www.far.ai/blog/far-ai-selected-to-lead-eu-ai-act-cbrn-risk-consortium;https://arxiv.org/pdf/2507.06260;https://arxiv.org/pdf/2601.19134,Not a TED tenderer (subcontractor only). No Coefficient or SFF hits.
EV09,FAR.AI,"nonprofit (501(c)(3) FAR AI Inc, EIN 92-0692207)",2022 (incorporated Oct 2022),US (Berkeley / San Diego registration),"AI safety research, red-teaming for frontier labs, regranting, events (Alignment Workshop), FAR.Labs coworking.","'Started red teaming leading language models for frontier labs' (2023, own timeline); paid consulting to frontier developers under its consulting policy; leads EU CBRN consortium (Lot 1). https://www.far.ai/about",59347676,18,6512577,"2024 990 revenue $23,254,628 (2023: $7,648,387; 2022: $1,507,686); Coefficient dominant incl. $28,675,000 (2025-09) and $12,000,000 regranting (2024); SFF 2022-H2 $524,000, 2023-H1 $2,028,000 + $1,063,000 (Labs), 2023-H2 $472,577 + $404,000, 2024 $383,000, 2025 $919,000 + $719,000 match; also fiscal host for Safe AI Forum ($580,000 SFF 2024, excluded)","yes, capped: 'Revenue from for-profit AI developers ... is capped at a maximum of 10% of FAR.AI's total annual revenue'; charges market rates to for-profit developers; retains right to publish","EU AI Office Lot 1 CBRN lead (EUR 1,434,080, with SecureBio and SaferAI; subcontractors GovAI, Nemesys, EquiStamp), awarded 2025-12-04; pro bono/discounted work for governments per policy","Adam Gleave, co-founder and CEO (UC Berkeley CHAI PhD; ex-Google DeepMind; also METR board member); Karl Berzins, co-founder and President",none found,https://www.far.ai/about;https://www.far.ai/transparency;https://www.far.ai/blog/far-ai-selected-to-lead-eu-ai-act-cbrn-risk-consortium;https://www.far.ai/team;https://projects.propublica.org/nonprofits/organizations/920692207;https://survivalandflourishing.fund/2025/recommendations;ted-864574-2025.xml,Coefficient name matched: 'FAR AI' (18 grants). Gleave sits on METR's board (board.csv B05): structural overlap.
EV10,SaferAI,nonprofit (French association loi 1901),2022,France (Paris),"Quantitative AI risk models, company risk-management ratings, standards work (OECD, NIST, ISO, EU Code of Practice).",Partnered with Google DeepMind researchers on goal-directedness evaluations; external review of DeepMind's scheming-inability safety case (Jul 2026); advised G42 and 'a major AI company' on frontier safety frameworks; rates OpenAI/Anthropic/Google/Meta/xAI risk management publicly. https://www.safer-ai.org/about,0,0,1056800,"SFF 2022-H2 $214,000; 2023-H2 'Campos Safer AI Project' $325,800 via Lightspeed Grants (flagged); 2024 $95,000; 2025 $311,000 + $111,000 match; other funders not listed on site",unknown: no policy found on site; advisory work for G42 and 'a major AI company' suggests paid framework consulting,"EU AI Office Lot 1 consortium member (EUR 1,434,080 to consortium; leads risk modelling and monitoring workstreams, 3 years)","Henry Papadatos, Executive Director (ex-UC Berkeley sycophancy research); Simeon Campos, founder and board chair",none found,https://www.safer-ai.org/about;https://www.safer-ai.org/saferai-selected-to-support-eu-ai-act-implementation-through-cbrn-risk-assessment;https://survivalandflourishing.fund/2022/h2/recommendations;https://survivalandflourishing.fund/2025/recommendations;ted-864574-2025.xml,"No Coefficient hit under 'SaferAI'/'Safer AI'. The 2023-H2 'Campos Safer AI Project' row is attributed to SaferAI on the founder's name; excluding it gives $731,000."
EV11,UK AI Security Institute (AISI),government (DSIT),2023 (as AI Safety Institute; renamed Feb 2025),UK,"Government pre-deployment testing of frontier models, safety/security research, and grant programmes.","Pre-deployment access to leading models and 'close collaborations' with AI companies; joint work with OpenAI, Anthropic, Google DeepMind on model security and testing (own site). https://www.aisi.gov.uk/about",0,0,0,"UK government: 'GBP 66m in funding per financial year'; priority access to >GBP 1.5bn UK compute; Alignment Project GBP 27m total incl. GBP 5.6m from OpenAI plus Microsoft, Anthropic, AWS, Schmidt Sciences, Halcyon Futures, Safe AI Fund, CIFAR, Canadian AISI, ARIA, UKRI, Renaissance Philanthropy, Sympatico Ventures, AI Safety Tactical Opportunities Fund","yes (for grants): Alignment Project coalition takes OpenAI (GBP 5.6m), Anthropic, Microsoft and AWS money; core budget is public","is the government; grants out: Alignment Project (60 projects, up to GBP 1m each; awardees include Jacob Hilton / Alignment Research Center; no METR, Apollo, Redwood, FAR, Epoch, Transluce, Palisade in the PDF), Challenge Fund (20 projects, up to GBP 200k), Systemic Safety grants; METR and EquiStamp report AISI work with undisclosed amounts","Henry de Zoete, Director (ex-Prime Minister's AI adviser, tech investor); Jade Leung, CTO (ex-OpenAI governance lead); Ian Hogarth, Chair",none found,https://www.aisi.gov.uk/about;https://www.aisi.gov.uk/grants;https://www.aisi.gov.uk/blog/funding-60-projects-to-advance-ai-alignment-research;https://alignmentproject.aisi.gov.uk/;evaluators/aisi-alignment-awardees.pdf,Awardee PDF lists names and institutions but no per-grant amounts. Challenge Fund awardee list not found.
EV12,"US Center for AI Standards and Innovation (CAISI, NIST)",government (Commerce/NIST),2023 as US AI Safety Institute; renamed CAISI June 2025,US,"Voluntary pre-deployment testing agreements with frontier labs, national-security capability evaluations, standards.","Agreements with OpenAI and Anthropic (Sep 2025) and Google DeepMind, Microsoft, xAI (May 2026); >40 evaluations incl. unreleased models; some classified via TRAINS taskforce. https://www.nist.gov/caisi",0,0,0,US appropriations: $10M for CAISI within $55M NIST AI line (FY2026); FY2027 request $27M ($11M base + $16M expansion),no (public budget); receives free pre-deployment model access from labs under voluntary agreements,"is the government; third-party evaluator agreement with Scale AI (Feb 2025, no amount); no CAISI contract to METR/Apollo/etc. found","Arvind Raman, NIST director and acting CAISI director since director Chris Fall resigned Jul 20 2026 (local independence_fight.csv IF47)",none found; no CAISI statement on designated evaluators Sep 9-13,https://www.nist.gov/caisi;https://ifp.org/funding-for-caisi/;https://ari.us/wp-content/uploads/2026/04/House-Letter-FY27-NIST-CAISI-Funding.pdf;https://labs.cloudsecurityalliance.org/research/csa-research-note-caisi-frontier-ai-testing-agreements-20260/;https://scale.com/blog/first-independent-model-evaluator-for-the-usaisi,"Budget figures from IFP / trade-association letters found by search, not from an appropriations text fetched here."
EV13,"EU AI Office (DG CNECT) - 'Technical Assistance for AI Safety' contractors, TED 864574-2025",government (European Commission),2024 (AI Office),Belgium (Brussels),"Enforces GPAI obligations under the AI Act; procures external risk modelling and evaluation capacity (six lots, EUR 9,330,000 estimated).","Lot winners (award dates Nov-Dec 2025): LOT-0001 CBRN EUR 1,434,080 -> FAR AI Inc (lead), SecureBio, SaferAI [10 bids]; LOT-0002 Cyber Offence EUR 1,700,000 -> Capgemini Consulting (lead), Wavestone, LNE [6 bids]; LOT-0003 Loss of Control EUR 1,167,484 -> EquiStamp Inc (lead), Model Evaluation and Threat Research Inc (METR), Epoch Artificial Intelligence Inc [7 bids]; LOT-0004 Harmful Manipulation EUR 881,345 -> EquiStamp Inc (lead), Clarity AI Research Inc (Transluce), Centre pour la Securite de l'IA (CeSIA) [12 bids]; LOT-0005 Sociotechnical EUR 1,221,587.50 -> Technopolis Consulting Group Belgium (lead), WIK-Consult, TNO [13 bids]; LOT-0006 Agentic Evaluation Interface EUR 968,521 -> French Ministere du Budget (lead) with LNE [15 bids]. ted-864574-2025.xml",0,0,0,"EU budget, Digital Europe Programme (FundingProgramCode DIGITALEU)",no (public budget),"is the contracting authority; total awarded EUR 7,373,017.50 across six lots",not on the fetched AI Office page (leadership names not recorded),none found,ted-864574-2025.xml;https://digital-strategy.ec.europa.eu/en/policies/ai-office;https://www.far.ai/blog/far-ai-selected-to-lead-eu-ai-act-cbrn-risk-consortium,"Notice ID 90db1f37-d112-4021-a742-1a4db97f1c43, issued 2025-12-23. Subcontracting 'yes' on lots 1, 3, 4, 5. GovAI and Nemesys (Lot 1) and Apart Research (Lot 4, per a search snippet) are subcontractors, not tenderers."
EV14,Stanford NLP Group (Stanford University),academic,n/a (Stanford NLP group; Manning director of SAIL),US,University NLP research group; proposed by its director as a university-based third-party evaluator.,No lab evaluation engagement found; Percy Liang (Stanford NLP/CRFM faculty) advises Transluce and holds Coefficient grants for LM evaluation and cyber benchmarks. https://x.com/chrmanning/status/2098904071407362351,10126544,9,0,"Stanford University overall in Coefficient index: $53,364,063 across 52 grants (mostly non-NLP: biomedical, lead, SERI); NLP-faculty subset only counted here; university and industry gifts not enumerated",unknown for the group; Stanford as a university takes industry gifts; no group-level policy,"none specific found (Stanford PIs Silverstein, Duchi, Van Roy hold UK AISI Alignment Project grants, but not NLP group)","Christopher Manning, Director of Stanford NLP and SAIL (Thomas M. Siebel Professor); Percy Liang, Stanford NLP / CRFM director","Christopher Manning Sep 12 22:38 UTC: 'I propose Stanford NLP as an independent third-party evaluator under @DarioAmodei's 3 step plan' (240,952 views)",https://x.com/chrmanning/status/2098904071407362351;https://api.fxtwitter.com/i/status/2098904071407362351;https://www-nlp.stanford.edu/~manning/;coefficient index 'Stanford University',"AMBIGUOUS Coefficient match: 'Stanford University' covers the whole university. Counted only NLP-adjacent PIs named in grant titles: Percy Liang ($1,337,600 2017; $25,000 2017; $1,500,000 2021; $250,000 2023 LM evaluations; $2,937,000 2024 cyber benchmark), Tatsunori Hashimoto ($880,000 2024; $1,922,145 2025), Christopher Potts ($743,500 2025), Diyi Yang ($531,299 2025). No grant names Manning. Thread continuation beyond the first post not retrievable via fxtwitter. audit 2026-09-14: institution-wide count was 53 grants; exact organization_name matching in the local Coefficient index yields 52."
EV15,Hugging Face - Open Alignment Initiative,company (VC-backed; Nvidia acquisition reported Aug 2026),2016 (company); initiative announced Sep 10-12 2026,US / France,New alignment/safety team for open models inside Hugging Face; asked to join Anthropic's embedded-evaluator programme.,None as evaluator yet; Delangue asked Sep 12 to be part of Anthropic's programme; HF was the victim platform in the OpenAI agent incident (Aug 2026). https://x.com/ClementDelangue/status/2098790988034580852,0,0,0,"Series D $235M (Aug 2023, $4.5B valuation) led by Salesforce Ventures with Google, Amazon, Nvidia, AMD, Intel, IBM, Qualcomm, Sound Ventures; Nvidia reported to have agreed to acquire HF for ~$12.9-13B (TechCrunch Aug 26 2026, Forbes Aug 27 2026)","yes: investors include Google, Amazon and Nvidia; pending Nvidia ownership; no evaluator-independence policy published",none found,"Clement Delangue, co-founder and CEO; Thomas Wolf, co-founder and Chief Science Officer (leads Open Alignment)","Clement Delangue Sep 12 15:08 UTC (378,966 views): launches Open Alignment Initiative led by Thomas Wolf and asks to join the embedded-evaluators programme; Wolf Sep 12 16:24 backs third-party evaluators 'if done right'",https://x.com/ClementDelangue/status/2098790988034580852;https://x.com/Thom_Wolf/status/2098810070423216324;https://www.techmeme.com/260912/p13;https://techcrunch.com/2023/08/24/hugging-face-raises-235m-from-investors-including-salesforce-and-nvidia;https://techcrunch.com/2026/08/26/nvidia-closes-in-on-hugging-face-acquisition/,No HF blog post for the initiative found (huggingface.co/blog/open-alignment 404); Dealroom note (Sep 10 Wolf post) blocked by JS wall. Nvidia acquisition status as of Sep 13 rests on press reports.
EV16,"Scale AI - SEAL (Safety, Evaluations and Alignment Lab)",company,2016 (Scale); SEAL Nov 2023,US (San Francisco),Commercial evaluation and red-teaming arm of Scale AI; SEAL leaderboards; Humanity's Last Exam with CAIS.,'Leading test and evaluation partner for frontier AI companies' (own blog); paid evaluations for frontier developers; first third-party evaluator agreement with US AISI/CAISI (Feb 2025). https://scale.com/blog/safety-evaluations-alignment-lab,0,0,0,Meta bought 49% (non-voting) for $14.3B in June 2025 ($29B valuation); earlier VC rounds,yes: revenue model is paid evaluation/data services to frontier labs; 49% owned by Meta (a frontier developer),US AISI/CAISI evaluator agreement Feb 2025 (no amount); Scale also sells to US DoD (Donovan) - amounts not researched,"Francis deSouza, CEO (appointed 2026 per scale.com banner); Alexandr Wang, founder, left for Meta June 2025; Summer Yue, SEAL Director of Research (Feb 2025)",none found,https://scale.com/blog/safety-evaluations-alignment-lab;https://scale.com/blog/first-independent-model-evaluator-for-the-usaisi;https://www.cnbc.com/2025/06/12/scale-ai-founder-wang-announces-exit-for-meta-part-of-14-billion-deal.html;https://techcrunch.com/2025/06/13/scale-ai-confirms-significant-investment-from-meta-says-ceo-alexandr-wang-is-leaving/,scale.com/seal returns 404; SEAL now lives at scale.com/leaderboard and blog posts.
EV17,MLCommons (AILuminate),"nonprofit (501(c)(6) industry consortium, EIN 85-0546914)","2020 (from MLPerf, 2018)",US (Delaware registration),Industry benchmark consortium; AILuminate safety/jailbreak/agentic benchmarks and Global Assurance Program.,"AI Risk & Reliability working group contributors include Anthropic, OpenAI, Google, Meta, Microsoft, Nvidia, Cohere; AILuminate Global Assurance backed by KPMG, Google, Microsoft, Qualcomm. https://mlcommons.org/ailuminate/",0,0,0,"125+ founding members and affiliates (dues); ProPublica shows net assets $4,551,326 (latest) with revenue fields blank/zero in the summary",yes: member-funded consortium whose members include the frontier labs,none found (collaborates with AI Verify Foundation and LNE),"Peter Mattson, President and founder (Google senior staff engineer); David Kanter, founder and Head of MLPerf",none found,https://mlcommons.org/about-us/;https://mlcommons.org/about-us/leadership/;https://mlcommons.org/2026/02/ailuminate-global-assurance/;https://arxiv.org/abs/2503.05731;https://projects.propublica.org/nonprofits/organizations/850546914,"ProPublica summary rows show 'Total Revenue $0' beside multi-million net assets; treat revenue as unrecorded, not zero."
EV18,RAND Corporation - Canary (with METR),"nonprofit (501(c)(3), EIN 95-1958142; FFRDC operator)",1948 (RAND); Canary 2024,US (Santa Monica),"RAND-METR collaboration building pre-deployment evaluation science, especially AIxBio misuse evaluations.",Canary publishes AIxBio agent evaluations (2026); METR side carries the lab engagements; no RAND-lab evaluation contract found. https://www.rand.org/global-and-emerging-risks/centers/ai-security-and-technology/projects/canary.html,86955751,17,1044000,"Audacious Project ~$38M Canary commitment (~$21M RAND side, 2024-10); Coefficient $10,000,000 'AI Evaluation and Testing' (2025-09-20) among $87M to RAND since 2020; FY2024 990 revenue $488,777,692 (mostly US-government contract research); SFF 2025 $1,022,000 + $22,000 match to the Technology and Security Policy Center (not Canary)",unknown: RAND has no published policy on frontier-lab funding; none identified,RAND operates federally funded R&D centers for DoD/DHS; no CAISI/AISI/EU evaluation contract found,"Jason Matheny, President and CEO (ex-White House OSTP/NSC, ex-IARPA)",none found,https://www.rand.org/global-and-emerging-risks/centers/ai-security-and-technology/projects/canary.html;https://www.rand.org/news/press/2024/10/09.html;https://projects.propublica.org/nonprofits/organizations/951958142;https://survivalandflourishing.fund/2025/recommendations,"Coefficient name matched: 'RAND Corporation' (17 grants; Bouisson-Bertrand Institute excluded). Grants are institution-wide (biosecurity, TASP, Meselson Center), not Canary-specific except the $10M 2025-09-20 line."
EV19,Palisade Research,"nonprofit (501(c)(3) Palisade Research Inc, EIN 93-1591014)",2022-2023,US (Berkeley; DE registration),"Studies strategic/dangerous capabilities of released models (shutdown resistance, hacking, cheating); briefs Congress.","Post-release research on OpenAI o3 and other models (shutdown resistance, 2025); no pre-deployment lab engagement found; work cited by Bengio and Amodei per own site. https://palisaderesearch.org/about",3803463,2,3212627,"2024 990 revenue $2,887,428 (2023: $669,000); SFF 2023-H2 $401,627 via LSG + $164,000; 2024 $586,000; 2025 $1,133,000 + $928,000 match; donors not named on site",unknown: no policy on site,"none found (briefings to Congress/White House staff, unpaid)","Jeffrey Ladish, founder and Executive Director (ex-Anthropic security team, 2022)",none found,https://palisaderesearch.org/about;https://palisaderesearch.org/donate;https://projects.propublica.org/nonprofits/organizations/931591014;https://survivalandflourishing.fund/2023/h2/recommendations;https://survivalandflourishing.fund/2024/recommendations;https://survivalandflourishing.fund/2025/recommendations,"Coefficient name matched: 'Palisade Research' ($1,680,000 2024-06; $2,123,463 2025-05)."
EV20,Alignment Research Center (ARC),"nonprofit (501(c)(3), EIN 86-3605182)",2021,"US (Covina, CA registration)","Theoretical alignment research (mechanistic estimation); former parent of ARC Evals, now METR.","As ARC Evals: GPT-4 red-team access (2023), Claude 2 (2023); no evaluation work since the 2024 spin-out; Paul Christiano joined OpenAI Foundation board and Safety and Security Committee Sep 9 2026. https://openai.com/index/paul-christiano-joins-openai-foundation-board/",1515000,2,2376000,"SFF 2022-H2 $2,179,000; 2024 $197,000; Effective Ventures USA $1,500,000 (FY2023, snippet only); 990 revenue $475,650 (2021), $4,466,000 (2022), $10,043,543 (2023), $5,219,333 (2024); transferred $4,553,935 to METR in 2024",unknown: no policy found; Christiano now holds OpenAI Foundation roles,"UK AISI Alignment Project grant to Jacob Hilton / ARC (2026, amount not listed; up to GBP 1m per project); none other","Paul Christiano, Executive Director and President (ex-OpenAI alignment lead; US CAISI head of AI safety 2025; OpenAI Foundation board Sep 2026); Jacob Hilton, researcher and 2024 990 President (ex-OpenAI)",none found; Christiano's OpenAI Foundation appointment Sep 9 is not a proposal of ARC as evaluator,https://www.alignment.org/;https://www.alignment.org/team/;https://projects.propublica.org/nonprofits/organizations/863605182;https://survivalandflourishing.fund/2022/h2/recommendations;https://survivalandflourishing.fund/2024/recommendations;evaluators/aisi-alignment-awardees.pdf;https://techcrunch.com/2026/09/09/openai-adds-a-prominent-ai-doomer-to-its-board-of-directors/,"Coefficient name matched: 'Alignment Research Center' ($265,000 2022-03; $1,250,000 2022-11). SFF 2023-H1 'ARC Evals Team' $3,247,000 is counted on the METR row, not here. Team page says Christiano is executive director (Apr 2024 page). audit 2026-09-14: named_in_sep_2026 was Christiano's OpenAI appointment, not a proposal of ARC as evaluator."
EV21,Halcyon Futures,"nonprofit incubator and grant fund (Halcyon Futures Inc, EIN 93-3537739) plus VC fund (Halcyon Ventures)",2023-09,US (San Francisco),"Seeds and incubates AI safety, biosecurity and cybersecurity orgs; Request for Founders lists ""Independent AI evaluators, auditors and accountability mechanisms: Rigorous third-party testing of frontier AI models before and after deployment""","Not an evaluator; funder of evaluator start-ups. Site: ""A deep network at frontier AI labs""; incubated AIUC (Rajiv Dattani, ex-METR COO). https://halcyonfutures.org/founders",0,0,0,"self-reported $25M raised for nonprofit and VC fund (McCormick X, 24 Sep 2025), funders unnamed; absent from SFF round pages 2019–2025 and from the Coefficient index; 990 FY2024 grants paid $611,455 (M125)",unknown: no lab money policy published,none found,"Mike McCormick, CEO (ex-GPV partner); Ross Matican, Investor & Grantmaker (ex-Bessemer); Shelby Summerfield",yes: McCormick's 12 Sep 2026 offer to fund eval and audit orgs (IF93),https://halcyonfutures.org/;https://halcyonfutures.org/founders;https://x.com/MikeMcCormick_/status/1970932733485646216;https://www.matsprogram.org/stream/halcyon;https://projects.propublica.org/nonprofits/organizations/933537739,Co-mentors a MATS Winter 2027 Founding & Field-Building stream (public founder pipeline). S4-people gap 4
EV22,Artificial Intelligence Underwriting Company (AIUC),for-profit (AI agent standards; accredited-auditor certification; liability insurance),2025-07,US (San Francisco),"Publishes the AIUC-1 standard (""SOC-2 for AI agents""), certifies AI vendors against it via accredited auditors (Schellman, Coalfire), and sells liability insurance priced on audit results; certified customers listed include Cursor, ElevenLabs, Harvey, KPMG","None as evaluator of frontier labs; AIUC-1 C010 (mandatory) requires certified companies to ""Appoint expert third parties to evaluate system robustness ... at least every 3 months""",0,0,not checked,"$15M seed 2025-07-23: Nat Friedman/NFDG (lead), Emergence, Terrain, Ben Mann (Anthropic co-founder), ex-CISOs of Google Cloud and MongoDB; Halcyon Futures career grant to Dattani and investment in every round (amounts unstated)",seed investor Ben Mann is an Anthropic co-founder; no lab is a disclosed shareholder,none found,"Rune Kvist, co-founder and CEO (ex-Anthropic, CAIS board); Rajiv Dattani, co-founder (METR board member, ex-METR COO); Brandon Wang, co-founder and CTO",no,https://www.aiuc.com/team;https://www.prnewswire.com/news-releases/the-artificial-intelligence-underwriting-company-launches-with-15m-to-help-enterprises-deploy-ai-with-confidence-302512447.html;https://aiuc.com/research/introducing-aiuc-1;https://standard.aiuc-1.com/safety;https://halcyonfutures.org/,No AIUC page names METR as a tester. S9 gap 1
EV23,FAR.AI (FAR AI Inc; EIN 92-0692207),"nonprofit (501(c)(3)) research institute, events, FAR.Labs coworking, regranting",2022,US (Berkeley; registered San Diego),"Technical AI safety research (adversarial robustness, red-teaming, evaluations), workshops, regranting ($1.46M subgrants 2024), FAR.Labs coworking","GPT-5 System Card §6.3.3.4: ""FAR.AI conducted 80 hours of red-teaming over 1 week against the OpenAI API""; site: ""red-teaming engagements for leading AI companies""",59347676,18,not checked,Schmidt Sciences; Survival and Flourishing Fund; CSET; Frontier Model Forum's AI Safety Fund (>$30M 2025 commitments in total),"yes, indirect: AISF is supported by Anthropic, Google, Microsoft and OpenAI (FMF page); no cap policy published on far.ai; 990 TY2024 earned revenue $429,091 of $24,293,765",EU AI Office consortium work per Founders Pledge; not itemised,"Adam Gleave, co-founder and CEO (METR board member; 990 comp $229,331); Karl Berzins, co-founder and President/COO; board 3 voting, 2 independent",no,https://www.far.ai/about;https://www.far.ai/news/30m-multi-funder-support;https://www.frontiermodelforum.org/ai-safety-fund/;https://arxiv.org/pdf/2601.03267;https://gt990datalake-rawdata.s3.amazonaws.com/EfileData/XmlFiles/202543189349316054_public.xml;/mnt/f/projects/memes/anthropic-investors/research/04-openphil-grants-raw-2026-09-11.json,"Coefficient total excludes Safe AI Forum $1,267,513 (far.ai/about/saif). S9 gap 3"
EV24,LMArena (Arena),for-profit (model-evaluation leaderboard platform),2025-05 (company; UC Berkeley project since 2023),US (San Francisco),"Community-vote leaderboard ""for evaluating the best AI models""; labs list flagship models for public comparison","OpenAI, Google, Anthropic models evaluated on the platform (press release)",0,0,not checked,"$100M seed 2025-05-21 led by a16z and UC Investments with Lightspeed, Laude Ventures, Felicis, Kleiner Perkins, The House Fund",n/a (VC-funded),none found,not recorded,no,https://www.prnewswire.com/news-releases/lmarena-secures-100m-in-seed-funding-to-bring-scientific-rigor-to-ai-reliability-302462025.html,"Listed because a16z (former firm of METR advisor Marco Mascorro; partner Justine Moore amplified an anti-METR summary, IF123) co-led its seed. S9 gap 2"