Researchers can disable AI model training in many popular AI services. Doing so can help maintain your privacy and avoid violations of license agreements with content providers.
Below are methods for doing so with some of the top generative AI services.
ChatGPT
As of 2025, ChatGPT currently trains its models on user inputs, like prompts and file uploads.
Once logged in to your OpenAI account, follow these steps.
- Click on your account name in the lower left corner of the screen.
- Select settings
- Under Settings, select Data controls
- Under Improve the model for everyone, toggle this to Off, and select Done.
Microsoft Copilot
As of 2025, Microsoft has removed the ability for individuals to turn off model training.
Since Copilot uses a version of ChatGPT, the work around is to use a regular ChatGPT account through OpenAI instead and following the instructions noted above.
Google Gemini
Individual Google account owners can revoke permissions for Google to use your data for AI training. See below for information on organizational accounts.
Follow these steps for your personal account.
- Go to Settings and help on the Gemini chat interface and select Activity
- Under Keep Activity, set this to Off
For organizational accounts, such as university work accounts, control over model training in Google Gemini is reserved for organizational IT departments.
Organizational users wishing to work with Gemini without model training are advised to consult their campus IT personnel to turn off model training across all campus accounts.
Alternatively, organizational users can switch to a personal account and disable model training for that account.
Claude.ai
As of 2025, Antropic's Claude service trains its model on user data by default.
Follow these steps to turn off model training.
- In the lower left of the main chat screen, select your user account icon.
- Select Settings
- In Settings, select Privacy
- On the Privacy screen, toggle off Help improve Claude
Comments
0 comments
Please sign in to leave a comment.