Is It Safe to Put Personal Information in ChatGPT?

ChatGPT feels private. It is a one-to-one chat window, and it answers like a helpful colleague. But anything you type into it is data you have handed to a company, and the safety of that data depends on settings most people never open. The short answer is: ChatGPT is reasonably safe for everyday, non-sensitive use if you change two settings and follow a few habits — and it is not the place for passwords, financial account numbers, medical records, or other people’s private details.

This article explains what OpenAI says it does with your chats, which controls actually protect you, and what to do instead when the information is genuinely sensitive. Prices and model names are not the issue here; your data is.

What OpenAI collects when you use ChatGPT

OpenAI’s privacy policy, updated on 10 September 2026, says the company may collect the content you type, upload or speak into ChatGPT, along with account details and usage information. That is the starting point: assume anything you paste could be stored on OpenAI’s systems, seen by a limited number of authorised staff or trusted service providers for purposes such as investigating abuse or handling legal matters, and kept for a period after you delete it.

That last point matters. Deleting a chat removes it from your history immediately, but OpenAI schedules deleted chats for permanent deletion from its systems within 30 days, unless the data was already de-identified and separated from your account or must be retained longer for security or legal reasons. Deleting is real, but it is not instant, and it is not absolute.

The key question: does ChatGPT train on your chats?

For personal accounts — Free, Go, Plus and Pro — the answer, by default, is yes. OpenAI’s help centre says it may use content from those accounts to train its models, and that training is enabled by default. Paying for Plus or Pro does not switch it off.

Business, Enterprise and Edu workspaces are different: OpenAI says it does not train on their content by default. The same applies to the API. So the same words pasted into a personal account and a work account can be treated differently, and if your employer provides ChatGPT through a business workspace, your work chats have stronger default protection than your personal ones.

The one setting that stops training

The setting is called “Improve the model for everyone.” It lives under Settings, then Data controls, on the web and in the apps. Turn it off, and OpenAI says new conversations and tasks will not be used to train its models. Your existing chats stay in your history — turning the setting off does not delete or hide them.

You can also choose “Do not train on my content” in OpenAI’s Privacy Portal, without signing in, and OpenAI says either route is enough. If you use ChatGPT signed out, the choice is saved in that browser only, so it does not follow you to your phone or a private window.

There is a catch that catches many people out. If you opt out of training and then give a reply a thumbs up or a thumbs down, OpenAI says the entire conversation attached to that feedback may be used for training, even if you opted out. So if you have turned training off, do not rate replies in chats that contain anything sensitive.

Voice mode has its own training control, separate from the text setting. If you dictate to ChatGPT, check that control too.

Memory is a different switch — turn-off training does not stop it

Memory and training are two different things, and confusing them is the most common privacy mistake with ChatGPT.

Memory is what ChatGPT remembers about you across chats: details you ask it to remember, or things it saves because they seem useful, plus what it can draw from your past chats. The controls are under Settings, then Personalization, then Memory. You can view or remove individual saved memories, and you can choose “Delete and turn off memory” to clear what it remembers and switch memory off. OpenAI says that does not delete your past chats.

Turning off “Improve the model for everyone” does not stop ChatGPT remembering things about you. Turning off memory does not stop training. If you care about privacy, check both switches, not one.

Temporary Chat: the safest way to ask a sensitive question

Temporary Chat is ChatGPT’s private-chat mode. A temporary chat is not saved in your history, is not used to train OpenAI’s models while it remains temporary, does not create or update memories, and is deleted from OpenAI’s systems after 30 days. OpenAI says temporary chats may still be reviewed to monitor for abuse.

Use Temporary Chat for anything you would not want sitting in your history: a health question, a financial worry, a draft message about a personal conflict. It is the closest ChatGPT comes to a conversation that leaves no trace in your account — while remembering that OpenAI may keep a copy for up to 30 days for safety purposes.

What never to paste into ChatGPT

Some information should not go into any consumer chatbot, whatever the settings say. Do not paste passwords, bank account or card numbers, national insurance or Social Security numbers, passport or driving licence numbers, or full medical records. Do not paste other people’s private information — a colleague’s salary, a friend’s health details, a child’s school records — without their clear consent. And do not paste confidential work documents, client data or trade secrets into a personal account, where training is on by default.

If you need help with a document that contains sensitive details, remove the identifying parts first — ask about a contract clause with names and figures taken out, or a medical question in general terms, not by uploading the full document or test results.

How to export and delete your data

You have the right to a copy of your data and to ask for it to be deleted, and OpenAI provides both routes in Settings under Data controls, or through its Privacy Portal.

To export, go to Settings, then Data controls, then Export data, and confirm the export. OpenAI says the export can take up to seven days to prepare, and the download link expires after 24 hours, so download it promptly when the email or text arrives. Exporting creates a copy; it does not delete anything.

To delete a single chat, use the menu beside it in your history. To delete every conversation, use Settings, then Data controls, then Delete all chats. Deleted chats disappear from view immediately and are scheduled for permanent deletion within 30 days, subject to the security, legal and de-identification exceptions described above.

To delete your whole account, use Settings, then Account, then Delete account, or submit the request through the Privacy Portal. Account deletion is permanent and cannot be undone, so export anything you need first.

A safer way to use ChatGPT: five habits

First, turn off “Improve the model for everyone” today, and check the memory setting while you are there. Second, use Temporary Chat for anything personal. Third, never paste passwords, account numbers, ID numbers or full medical records. Fourth, strip names, addresses and figures out of documents before you ask for help with them. Fifth, review your saved memories and chat history every few months and delete what you no longer need — and remember that a thumbs up or down on a reply can put that whole conversation back into training, even after you opted out.

ChatGPT is a powerful tool, and for most everyday questions it is safe enough. The risk is not that it is spying on you; it is that its defaults favour training over privacy, and that most people never change them. Change them, keep the most sensitive details out entirely, and you keep the benefit while cutting most of the risk. For the wider comparison of what the three big assistants cost and offer, read ChatGPT vs Claude vs Gemini: Which AI Should You Use in 2026?, and for help judging claims you see online, read How to Check a Viral News Claim Before You Share It.

Frequently asked questions

Does ChatGPT use my chats to train its models?
On personal Free, Go, Plus and Pro accounts, yes, by default. Turn off “Improve the model for everyone” under Settings, then Data controls, to stop new conversations being used for training. Business, Enterprise and Edu workspaces do not train on content by default.

Does paying for ChatGPT Plus make my data private?
No. Paying does not switch off training. Plus and Pro users need to turn off the same setting as free users. The paid plans remove ads and raise usage limits; they do not change the training default.

What is Temporary Chat in ChatGPT?
A chat that is not saved in your history, is not used for training while it remains temporary, and does not create memories. It is deleted from OpenAI’s systems after 30 days, and may be reviewed for abuse monitoring.

Can I delete everything ChatGPT knows about me?
You can delete individual chats, delete all chats, clear saved memories, and delete your whole account. Deleted chats are scheduled for permanent deletion within 30 days, with limited exceptions for de-identified data and security or legal retention. Export anything you need before deleting, because deletion cannot be undone.