Prompt & Done · PRIVACY
The 3 off switches for AI training on your chats
ChatGPT, Gemini and Claude can use your chats to train their models, on free and paid personal plans. Here's where each switch is, and the one catch.
Checked 5 Nov 2026 on each company's own help pages. Menus move; the source links are below each step.
1 ChatGPT
Settings → Data controls → Improve the model for everyone → off → Done. On mobile: sidebar → your profile → Data controls. It's on by default for Free, Plus and Pro, and turning it off keeps your chat history.
Codex has its own separate setting (Codex settings → Include environments).
2 Gemini
Turn off Keep Activity (in Gemini: Settings → Activity, or at myactivity.google.com/product/gemini). With it on, Google uses your activity to improve its services, including training generative AI models, with help from human reviewers.
The trade-off: with it off, your chats aren't saved to your activity (Google keeps them 72 hours to respond and for safety).
3 Claude
Settings → Privacy → Help improve our AI models → off. Applies to Claude Free, Pro and Max (and Claude Code on those plans). New chats and coding sessions stop being used for training.
! The catch: 👍 and 👎
Even with training off, rating a reply can send that conversation in. OpenAI: "if you submit feedback … we may use the entire conversation associated with that feedback to improve our models, even if you have opted out." Gemini: with Keep Activity off, feedback sends the feedback plus your last 24 hours of chats.
For a one-off sensitive question, use a temporary chat (ChatGPT, Gemini) or an incognito chat (Claude).
Work accounts
ChatGPT Business/Enterprise/Edu, Claude for Work and the API, and Google Workspace Gemini have different defaults set by your organisation.