# Meta AI privacy rating

AI assistant from Meta built into Facebook, Instagram, WhatsApp and Messenger, and available on the web and as a standalone app.

## Summary

Meta AI scores 27 out of 100 (grade F) on the AI assistants criteria. It meets 3 of 12 criteria: transparency report, tells users about requests and TLS configuration. It partly meets limited chat retention and no account needed. It does not meet open source, no trackers or telemetry, no ads or data sales, independent audit, security headers, no training on your data and runs locally. It is based in the United States: Five Eyes member; subject to the US CLOUD Act. Automated tests: SSL Labs grade A and Mozilla HTTP Observatory grade C.

- Website: https://www.meta.ai
- Jurisdiction: United States. Five Eyes member. Subject to the US CLOUD Act.
- Platforms: Web, Android, iOS
- Category: [AI assistants](https://privacyratings.com/ai-assistants/)
- Also rated: [Facebook Messenger](https://privacyratings.com/messengers/facebook-messenger/) (Messengers), [WhatsApp](https://privacyratings.com/messengers/whatsapp/) (Messengers), [Facebook](https://privacyratings.com/social-networks/facebook/) (Social networks), [Instagram](https://privacyratings.com/social-networks/instagram/) (Social networks), [Threads](https://privacyratings.com/social-networks/threads/) (Social networks)
- Grade: F (27/100)

## Criteria

| Criterion | Answer | Note | Evidence |
| --- | --- | --- | --- |
| Open source | No | Closed source. Meta publishes some Llama model weights, but the Meta AI service is not open source. |  |
| No trackers or telemetry | No | Meta collects activity and device data across its products, including Meta AI, and uses it for personalization and ads. | https://www.facebook.com/privacy/policy/ |
| No ads or data sales | No | Interactions with Meta AI are used to personalize ads across Meta's apps. | https://about.fb.com/news/2025/10/improving-your-recommendations-apps-ai-meta/ |
| Independent audit | No | No independent audit is published. |  |
| Transparency report | Yes | Publishes counts of government requests for user data twice a year. | https://transparency.meta.com/reports/government-data-requests/ |
| Tells users about requests | Yes | Meta's policy is to notify people of requests for their information before disclosure, unless prohibited by law or in exceptional circumstances. | https://www.facebook.com/safety/groups/law/guidelines/ |
| TLS configuration | Yes | Grade A | https://www.ssllabs.com/ssltest/analyze.html?d=www.meta.ai&hideResults=on |
| Security headers | No | Grade C (50/100+) | https://developer.mozilla.org/en-US/observatory/analyze?host=www.meta.ai |
| Modern web standards | Not tested yet | Not tested yet. |  |
| No training on your data | No | Interactions with AI features are used to train Meta's models, with no general opt-out setting. | https://www.facebook.com/privacy/genai/ |
| Runs locally | No | Hosted only. |  |
| Limited chat retention | Partial | Chats are kept until deleted. The privacy policy sets retention case by case, with no fixed deletion period. | https://www.facebook.com/privacy/policy/ |
| No account needed | Partial | The web version offers limited chat before logging in; most features need a Meta account. | https://www.meta.ai |

Source: https://privacyratings.com/ai-assistants/meta-ai/
Edit: https://github.com/privacyratings/privacyratings.com/edit/main/ratings/ai-assistants/meta-ai.md