You pasted something private into an artificial intelligence assistant and wish you had not. It happens often, and most of it can be put right. Work through the steps in order: the first two matter most, because deleting a conversation does not undo everything.
The short answer. Change anything that works like a key (a password, a code, an access key), call your bank if card or bank details went in, then delete the conversation and check the setting that decides whether chats help train the assistant.
Last reviewed 15 October 2026. The deletion details come from each company’s own help pages, linked at the end, and they change, so check the page for the assistant you use.
1. Stop pasting, and work out what went in
Open the conversation and read what you pasted. Find the line below that matches. If more than one matches, do each one.
- A password, a one-time code, a recovery phrase or an access key. Change it now, on the real website or in the real app (type the address yourself). If it is a one-time code that was already used, there is nothing more to do. Change the same password anywhere else you used it.
- A card number or bank details. Call your bank on the number printed on the back of your card, say what you pasted and when, and ask what they advise. The bank is the one that can watch the account or replace the card.
- An identity number, such as a Social Security number, a passport number or a National Insurance number. In the United States, IdentityTheft.gov, run by the Federal Trade Commission, gives steps for the details you shared. In the United Kingdom, the National Cyber Security Centre lists what to do if you have shared sensitive information. Elsewhere, see official help, country by country.
- Someone else’s details, such as a customer, a patient, a student, a colleague or a relative. Tell the person who looks after data protection where you work, or your manager, today. In the United Kingdom, an organisation must report a notifiable personal data breach to the Information Commissioner’s Office “without undue delay, but not later than 72 hours after becoming aware of it”, which is why telling early matters. Pasting is not always a breach, and that is for the organisation to decide with the facts in front of it.
- A work document or work account details. Tell your information technology team or the person who looks after work devices. A question asked early is part of the job.
- Only your own name, address or phone number. Delete the conversation (step 2) and watch for unexpected messages. There is nothing to change.
2. Delete the conversation
Deleting removes it from your view straight away, and the company then deletes it from its own storage. What each company says today:
- Claude, from Anthropic. A deleted conversation disappears from your chat history immediately and is deleted from back-end storage within 30 days. Anthropic may keep chats where the law requires it, to resolve disputes, or to deal with breaches of its usage policy.
- ChatGPT, from OpenAI. A deleted chat disappears from your account view right away and is scheduled for permanent deletion within 30 days, unless it was already de-identified or OpenAI must keep it longer for security or legal reasons. Deleting a chat does not delete files saved separately in your Library, or the original data in an app you connected.
- Gemini, from Google. Open Settings & help, then Activity, and delete the item, or the last hour or day. Google says it first removes the activity from view and then starts a process to delete it from its storage systems, and does not give a timeline for the second step. With Keep Activity switched off, conversations are saved for up to 72 hours.
3. Check the training setting
Each assistant has a setting that decides whether your chats are used to improve its models. The name and the place differ, so search the help page for the assistant you use. On Gemini, the related setting is named Improve Google services with your audio and Gemini Live videos & screenshares, and Google says it is off by default. Switch it off if you do not want it on.
What deleting cannot do
Anthropic says deleted chats will not be used to train future models, but data from chats that were included while model improvement was on may still appear in training runs already in progress or in models already trained. That is why the first two steps matter most: anything that could be used to hurt you, such as a password or a card number, should be changed or cancelled, not just deleted.
Next time
Hide names before asking an artificial intelligence swaps names, phone numbers, email addresses and addresses in your message for labels such as [NAME 1], on your own device, so the assistant only ever sees the labels. What not to paste into artificial intelligence checks a message for details that should stay out. If you doubt it, UtilForge it.
Sources
- Anthropic Privacy Center, how long do you store my data (dated 1 July 2026)
- OpenAI Help, how to delete and archive chats in ChatGPT
- Google Gemini Help, manage and delete your Gemini Apps activity
- Information Commissioner’s Office, personal data breaches: a guide
- National Cyber Security Centre, if you have shared sensitive information
Related guides
Free
Hide the names
before you paste.
Hide Names works on your own phone or computer. What you type stays on your phone or computer.