On Claude Free, Pro and Max, open Settings, then Privacy, and switch off "Help improve our AI models." After that, new chats and coding sessions are not used for future training. The switch is not a time machine, and a few exceptions remain, so read the limits below.
Where the setting lives
Anthropic's privacy center describes the same path on every surface that has a settings menu.
On desktop or in a browser: click your name, choose Settings, choose Privacy, and find "Help improve our AI models." You can also go straight to claude.ai/settings/data-privacy-controls.
On mobile: tap your name in the settings menu, choose Settings, then Privacy, and flip the same toggle.
The setting applies to the consumer plans (Free, Pro, Max) and to Claude Code sessions run under those accounts. Team, Enterprise and API use is covered by separate commercial terms, so this toggle is not the control for those.
What turning it off does
Per Anthropic's help article, with the setting off:
- New chats and coding sessions will not be used for future model training.
- Chats and sessions you stored earlier will not be used in future training runs.
- Data that is already part of a training run that has started, or part of a model that is already trained, stays as it is.
That last point is the one people miss. Switching off does not pull your words out of a model that has already been trained.
What it does not cover
Anthropic lists a short set of cases where a conversation can still be used even with the toggle off.
Safety flags. If automated classifiers flag a conversation, it may still be used to improve internal trust and safety models, detect harmful content, enforce policies or advance safety research.
Feedback. If you press thumbs up or thumbs down, the related conversation is stored and may be used to train models where legally permitted. We cover that in a separate post on feedback buttons.
Explicit programs. Joining something like the Trusted Tester Program is an explicit opt-in to training.
A sensible routine
Check the toggle once after you create an account, because the default is a choice you should make on purpose. Check it again whenever a policy update email arrives. Anthropic keeps a public page of consumer terms and privacy policy updates, and a new notice is a good reason to look at your settings.
If you want a conversation that never feeds training regardless of the toggle, Anthropic says incognito chats are not used to improve Claude even when the setting is on. See our post comparing opt-out, deletion and incognito for how they differ.
Quick check
Does switching it off delete my old chats?
No. It stops them being used in future training runs. Deleting a chat is a separate action in your chat history.
Does the toggle remove my data from a model already trained?
No. Anthropic states that data already in training runs in progress, or in models already trained, remains included.
Sources
- Is my data used for model training? (Anthropic Privacy Center)
- How do I change my model improvement privacy settings? (Anthropic Privacy Center)
- How long do you store my data? (Anthropic Privacy Center)
Last reviewed: 19 October 2026.
