# AI Disclosure Copy Library

A library of approved UI copy for AI disclosures, per region and per surface. Treat as a single source of truth; product copy that differs from this library requires Legal + DPO sign-off.

## 1. Inline output labels

| Surface | Copy | When shown |
|---------|------|-------------|
| Output card badge | "AI-assisted" | every AI-produced output |
| Output card tooltip | "This was drafted by AI. Review before sending." | hover on badge |
| Abstain payload | "I cannot answer from the provided sources." | every abstain response |

## 2. First-use disclosure modals

### Global default

```
Title: This feature uses AI
Body: <Feature name> uses AI to <plain-language verb phrase>. AI can be wrong.
Always review outputs before relying on them.
[Link: Responsible AI Declaration]
Buttons: Got it. | Don't enable
```

### EEA-tailored

```
Title: This is an AI-assisted feature (EU AI Act)
Body: <Feature name> is an AI system within the meaning of EU Regulation 2024/1689.
You are interacting with an AI. AI outputs may be inaccurate.
We process your inputs through <provider> under our Data Processing Addendum.
You can disable AI features at any time in workspace settings.
[Link: Responsible AI Declaration] [Link: DPA] [Link: Privacy Policy]
Buttons: I understand | Don't enable
```

### Kenya / Nigeria / South Africa

```
Title: AI feature consent
Body: <Feature name> uses AI to <verb phrase>. By enabling it, you consent to your data being processed by our model providers under our DPA. We will not use your data to train general AI models.
[Link: DPA] [Link: Privacy Policy]
Buttons: I consent | Don't enable
```

## 3. High-risk decision oversight notice (per-step approval surface)

```
Title: A human must approve this step
Body: This action requires your approval before it runs.
Action: <action>
Inputs: <inputs>
Likely outcome: <outcome>
Buttons: Approve | Edit | Reject
```

## 4. Generative content disclosure (Art. 50)

```
Surface: at the top of any generated long-form artefact
Body: "This content was generated by AI."
```

## 5. Right-to-explanation copy

```
Title: How did the AI produce this?
Body:
- Inputs the AI saw: <list>
- Sources used: <citations>
- Model: <model name + version, prompt tag>
- Confidence: <if available>
Contact <support@> for further explanation.
```

## 6. Banned phrases

The following phrases shall not appear in AI disclosures or marketing copy:

- "fully autonomous"
- "human-level"
- "guaranteed accurate"
- "always safe"
- "100% reliable"
- "AI doctor / lawyer / accountant"
- "unbiased"

Replace with measurable, evidenced phrasing.

## 7. Translation policy

- Disclosures in EEA are required in the user's display language; baseline EN/DE/FR/ES/IT/NL/PT.
- Disclosures in KE/NG/ZA: EN baseline; Swahili and other locales added as supported.
- Translation supplier reviewed by Legal; copy reviewed by DPO.