First, don’t enter any private or confidential information into ChatGPT and similar tools. It’s possible that developers may review your entries to improve the next version of their model.
If you want to make sure your inputs aren't used to improve the model, you can turn off that feature in the settings of ChatGPT.
- Click on your name
- then Settings
- then Data controls
- and turn off the switch called “improve the model for everyone.”
Temporary chat
Another option is to use the feature called “temporary chat.” At the top of the page, click on the menu that says “ChatGPT,” and then click the option that says “temporary chat.” Then your chat won’t appear in your history and ChatGPT won’t save anything from your conversation.
You can do the same in other tools.
- In Perplexity’s settings, under Preferences, the switch is called “AI data retention.”
- In Google’s Gemini, look under Settings & help in the lower left, then click "Activity." At the top of the page that comes up, you can click the menu that says “turn off.”
- In Microsoft Copilot, if you log in with your NetID (not your personal email), you'll be in a version of Copilot that doesn't use your data for training. You can verify it when you see a shield icon in the corner near your name.
- Claude doesn’t use your inputs for training unless you opt in.
- NotebookLM: "We value your privacy and never use your personal data to train NotebookLM."
Learn more
- Temporary chat FAQ - Open AI
- Data Controls FAQ - Open AI
- Data privacy & security - Perplexity AI
- Is my data used for model training? - Anthropic
- Training is not the same as chatting: ChatGPT and other LLMs don’t remember everything you say - Simon Willison
This answer adapted under Creative Commons Attribution 4.0 International License from the University of Arizona Libraries.