How to stop ChatGPT, Claude, Gemini and Copilot from training on your data
By Rafael · Updated · 3 min read
The short answer
Each major AI assistant has a privacy setting for this. On personal accounts, turn off ChatGPT's "Improve the model for everyone", Claude's "Help improve Claude", Gemini's "Keep Activity" or Copilot's model training toggle. Business and work plans generally don't train on your data by default.
First: personal account or business plan?
The settings below are for personal accounts (free and individual paid plans). Business, team and enterprise plans from these providers generally don't use your data for training by default and are governed by business terms. For work information, an approved business plan is the better fix than a personal setting.
Menus change often. These steps are current as of October 2026; if something has moved, search the app's help center for the setting name.
ChatGPT
- Open Data ControlsOn the web, open Settings, then Data Controls. In the mobile app, it's in the side menu under your profile.
- Turn off "Improve the model for everyone"Your conversations stay in your history but won't be used to train ChatGPT. The setting applies to your whole account.
- Use Temporary Chat for one-offsTemporary Chats don't appear in your history and aren't used to train models.
ChatGPT Business, Enterprise and Edu don't use your data for training by default.
Claude
- Open Privacy settingsIn Claude (Free, Pro or Max), open Settings, then Privacy.
- Turn off "Help improve Claude"Future chats and coding sessions won't be used to train models. It doesn't remove data from training that has already happened.
Claude for Work (Team and Enterprise) and the Claude API fall under Anthropic's commercial terms, which don't use your inputs or outputs to train models by default. Conversations flagged for safety review may still be used for trust and safety work on any plan.
Google Gemini
- Open Gemini Apps ActivityIn Gemini, open Settings, then Activity, or go to myactivity.google.com/product/gemini.
- Turn off "Keep Activity"Future chats won't be used to train Google's models unless you send feedback. Google still keeps chats for up to 72 hours to respond and keep the service safe.
- Know about human reviewChats already reviewed by people are kept separately for up to three years, even if you delete your activity. Don't share anything you wouldn't want a reviewer to see.
Gemini in Google Workspace business and education editions is different: Google says your content isn't human reviewed or used for AI model training outside your organization without permission.
Microsoft Copilot
In the consumer Copilot app, open Settings, go to Privacy, and turn off model training. Copilot can still remember your conversations for personalization; Microsoft just won't use them to train its models.
Microsoft 365 Copilot used with a work or school account is covered by enterprise data protection: prompts and responses aren't used to train the underlying models.
What these settings don't do
- They don't stop the provider from storing your chats to run the service, keep it safe, or meet legal obligations.
- They don't make it OK to paste passwords, account numbers or other people's sensitive information.
- They apply to your account, so each person on your team has to set them. That's one reason a managed business plan is easier for organizations.
For more on choosing safe accounts and rules, see is it safe to put business data in AI? and what shadow AI is.
Common questions
- Does turning off training delete my past chats?
- No. It affects future chats. Delete conversations separately if you want them gone, and note that data already used in training can't be removed from trained models.
- Do paid personal plans like ChatGPT Plus or Claude Pro train on my data?
- They can, depending on your settings. Paid individual plans use the same settings as free ones; business and team plans are the ones with no-training defaults.
- Is the API safer than the chat app?
- Generally yes for training: the major providers' APIs fall under business terms that don't use your data for training by default. You still need to handle security and retention in your own app.