What you type into an artificial intelligence goes to the company that runs it, and that company’s own settings decide whether it is kept, read by reviewers or used to train its models. The settings differ between companies and change over time, so read the company’s own page.
Last reviewed 2 October 2026. Each paragraph below summarises one company’s own help page as it read on that day. Companies change these terms, so use the links to read the current wording. Work and college accounts often follow a separate agreement, which your information technology team can tell you about.
OpenAI (ChatGPT)
- By default, conversations can be used to improve OpenAI’s models.
- The switch is called Improve the model for everyone, under Settings, then Data controls. When it is off, new conversations are not used to train models. Saved chats stay saved.
- A temporary chat does not appear in your history and is not used to improve models. OpenAI says it may be kept for up to 30 days for safety.
- If you give a thumbs up or down after opting out, OpenAI says the whole conversation attached to that feedback may be used to train models.
OpenAI: Data controls frequently asked questions
Anthropic (Claude)
- Anthropic says it uses consumer chats to improve its models if you choose to allow it in your privacy settings, if a conversation is flagged for safety review, or if you join a programme such as its Trusted Tester Program.
- Incognito chats are not used to improve Claude, even when the model improvement setting is on.
- Feedback you send is stored for up to five years, according to its page.
Anthropic: is my data used for model training?
Google (Gemini)
- Google says a subset of chats is reviewed by human reviewers, including trained service providers, and that reviewers see them disconnected from your account.
- Activity is deleted automatically after 18 months by default. You can choose 3 or 36 months. Chats that reviewers have read are kept for up to three years.
- The setting is called Keep Activity in Gemini Apps Activity. When it is off, future chats are not used to train Google’s models unless you send feedback, and Google still keeps them for 72 hours to run the service and protect people.
Every other artificial intelligence
Microsoft Copilot, Meta Artificial Intelligence, Grok, DeepSeek, Qwen and apps that offer several artificial intelligences in one place each have their own page. Ask the same four questions of each: is it kept, for how long, can people read it, and is it used to train models. UtilForge has not checked those pages.
What this means before you paste
A setting protects you from training, and it does not take back what you already sent. Leave out passwords, card and bank numbers, identity numbers, health details, other people’s details, staff records and client secrets. What Not to Paste into artificial intelligence lists them and shows which ones are in a text you are about to send. When the artificial intelligence needs the rest of a message about real people, Hide Names from artificial intelligence swaps their details for labels first.
Related guides
Free
Check it first.
Then paste.
What Not to Paste into artificial intelligence runs on your own phone or computer. What you paste stays on your phone or computer.