ChatGPT & Codex Privacy Settings
A clear guide to AI training opt-out, memory, and data-sharing controls.
Why this exists
On 8 and 9 September 2026, OpenAI said an internal model had produced a proof toward the Navier-Stokes Millennium Prize problem. NYU mathematician Tristan Buckmaster and collaborator Levent Alpoge had spent months on related work and put drafts into OpenAI Codex. They asked whether those Codex sessions had been used to train the model that ran the proof.
OpenAI said it did not look up their prompts or proofs. In the same period it also said it could not rule out that de-identified data from their use of OpenAI products had helped improve its models. That gap is why this page exists: the training switches for ChatGPT and Codex are easy to miss, and they are not in one place.
It also exists to make every opt-out easy to find, so people can turn each one off without digging through scattered settings menus.
How to do it
- Sign in to ChatGPT in this browser first. If you are signed out, the settings link opens the home page instead of the control.
- Open a row below. It goes straight to the official OpenAI page.
- Turn that control off there.
What opting out does not change
Turning training off does not delete your chat history. Your past chats stay. New chats are not used to train.
If you tap thumbs up or thumbs down on a reply, OpenAI may still use that conversation for training. Skip feedback if you do not want that.
ChatGPT
Improve the model for everyone
Training on your chats. Audio and video sit on this same pane.
Do not train on my content
Privacy portal. Use this if you cannot reach the in-app toggle.
Marketing measurement and personalized marketing
Sign in first, or the hash shows the home page.
Ads personalization
Different pane from marketing privacy.
Memory
Turn memory off so past chats are not reused.