AI Privacy Settings Explained: Stop ChatGPT, Claude and Gemini Training on Your Data
Date Published
Every major AI assistant lets you stop your conversations from being used to train future models, and it takes about a minute per app. In ChatGPT it is a switch called Improve the model for everyone under Data Controls. In Gemini it is the Keep Activity setting on the Gemini Apps Activity page. In Claude it is the model improvement choice under Privacy settings. Turning these off does not make your chats invisible to the company, and it does not change what your employer's business plan already does. Here is what each setting actually controls.
What does turning off training actually change?
Whether your conversations become teaching material for the next model. It does not stop the company from processing your chats to answer you, storing them for a period, or reviewing them for safety and abuse. Those are separate policies. So the honest summary is: the training toggle protects you from the long term risk of your words surfacing in a future model, not from the company seeing them today. For anything genuinely confidential, the rule from our guide on using ChatGPT safely at work still stands: do not paste it in the first place.
How do I turn off training in ChatGPT?
Profile icon → Settings → Data Controls → switch off Improve the model for everyone. On mobile the same control sits under your profile in the sidebar. The setting syncs across your devices once changed. ChatGPT also offers Temporary Chat, which is not used for training and is removed after about 30 days, useful for one-off sensitive questions. OpenAI documents all of this in its Data Controls FAQ, including the note that limited safety monitoring can still apply in rare high risk cases.
How do I turn it off in Gemini?
Open Gemini Apps Activity and turn Keep Activity off. With it off, future chats are not used to train Google's models and do not appear in your activity. Two details matter, both from Google's own help page: chats are still kept for up to 72 hours so Gemini can respond and handle feedback, and conversations selected for human review can be retained for up to three years, disconnected from your account. Turning off activity is not the same as no one ever reading it.
How do I check the setting in Claude?
Settings → Privacy → the model improvement option. Anthropic asks consumer users to choose whether their chats can be used to improve Claude, so the state of that switch depends on what you picked when prompted. Check it rather than assume. Anthropic's privacy center article on training data also notes that thumbs up or down feedback is stored for up to five years, and that incognito chats are never used for model improvement regardless of the setting.
What about work accounts and business plans?
Business and enterprise tiers generally exclude your data from training by default. ChatGPT Team and Enterprise, Claude for Work, Gemini for Workspace, and the API products are sold to companies with contractual promises that customer data is not used to train models. That is one of the strongest arguments for using the account your employer provides instead of a personal one: the protection is in the contract, not in a toggle you might forget. If your company has no such plan, our guide on reviewing contracts safely with AI shows the redaction habit that keeps you safe on a personal account.
Which setting should I actually use?
Turn training off everywhere, use temporary or incognito chats for anything sensitive, and prefer the work account. Those three habits cover almost every realistic risk for an office worker. The one they do not cover is pasting something you should never have pasted, and no setting fixes that. Set the toggles once, then spend your attention on what goes into the box.
A privacy toggle protects the future. What you paste protects the present.
Take three minutes now: open each assistant you use, find the training switch, turn it off, and note where the temporary chat option lives. You will not have to think about it again.
Frequently asked questions
Does ChatGPT train on my conversations by default?
On consumer plans the Improve the model for everyone setting is on unless you turn it off under Data Controls. Team, Enterprise, and API usage are excluded from training by default.
If I turn off training, can the company still read my chats?
Yes, for responding to you, short term storage, and safety review. The toggle controls model training only. Use temporary or incognito modes and avoid pasting sensitive material for stronger protection.
Does Gemini delete my chats when Keep Activity is off?
Not immediately. Google keeps them for up to 72 hours to respond and process feedback, and chats picked for human review can be retained up to three years, separated from your account.
Does Claude use my chats for training?
Anthropic asks consumer users to choose. Check the model improvement setting under Privacy to see your current state. Incognito chats are never used for training either way.
Is my work account safer than my personal one?
Usually yes. Business tiers promise in the contract that customer data is not used for training, which is stronger than a personal toggle. Ask your IT team which plan you are on.
Sources
Yes, if you follow a few rules. What happens to what you type, five things never to paste, the 10-second anonymizing trick, and the settings worth two minutes.
Redact the parties, run a structured clause review, verify flags in the original, escalate what matters. Safe AI contract review for non-lawyers, with the prompt.