Privacy in the age of AI
Every OpenAI, Anthropic and Google plan: does it train on your data, can you opt out, and what gets kept. Every claim links to the provider's own policy. Last verified 2026-07-22.
| Plan | Trains by default | Opt-out | Retention (default) | ZDR |
|---|---|---|---|---|
| ChatGPT Free | Yes | Yes | Deleted chats purged in 30 days | — |
| ChatGPT Plus | Yes | Yes | Deleted chats purged in 30 days | — |
| ChatGPT Pro | Yes | Yes | Deleted chats purged in 30 days | — |
| ChatGPT Business | No | — | Admin-controlled · deleted in 30 days | No |
| ChatGPT Enterprise | No | — | Admin-controlled · deleted in 30 days | No |
| ChatGPT Edu | No | — | Admin-controlled · deleted in 30 days | No |
| OpenAI API | No | — | 30-day abuse-monitoring window | Yes |
| OpenAI API + ZDR | No | — | Zero at rest (eligible endpoints) | Yes |
| Claude Free | Yes | Yes | 30 days off · up to 5 years if training on | — |
| Claude Pro | Yes | Yes | 30 days off · up to 5 years if training on | — |
| Claude Max | Yes | Yes | 30 days off · up to 5 years if training on | — |
| Claude Team | No | — | While active · deleted purged in 30 days | No |
| Claude Enterprise | No | — | Admin-configurable (min 30 days) | Conditional |
| Claude API | No | — | ≤30 days backend | Yes |
| Claude API + ZDR | No | — | Zero at rest · not for Fable/Mythos | Yes |
| Gemini app (free) | Yes | History off only | 18-month auto-delete default | — |
| Google AI Plus | Yes | History off only | 18-month auto-delete default | — |
| Google AI Pro | Yes | History off only | 18-month auto-delete default | — |
| Google AI Ultra | Yes | History off only | 18-month auto-delete default | — |
| Gemini for Google Workspace | No | — | Admin-determined (90 days–indefinite) | No |
| Gemini API (unpaid) | Yes | No | No fixed window published | No |
| Gemini API (paid) | No | — | 55-day abuse log only | Yes |
| Vertex AI | No | — | 24h cache · 90 days if abuse-flagged | Yes |
The patterns that matter
- Upgrading doesn't change the defaults. Plus, Pro, Max and Ultra treat your data exactly like the free tier: same training default, same opt-out controls. On business plans and the APIs the default itself flips to no training.
- Not all opt-outs are equal. ChatGPT and Claude: one toggle stops training, your chats stay. Gemini: the only opt-out also turns off chat history.
- The feedback trap. Thumbs-up/down can re-license a conversation for training even after you opt out. Google's version sends your last 24 hours of chats with it.
- Deletion is not erasure. Human-reviewed Gemini chats live 3 years off-account; safety-flagged Claude content 2 years; OpenAI keeps a ~30-day abuse-monitoring floor.
- ZDR is a contract, not a checkbox. Approval-gated, API-only, stateful features excluded. It also never covers Anthropic's newest models: Claude Fable 5 and Mythos 5 require 30-day retention even under ZDR. On Vertex it takes extra configuration to be real.