Is ChatGPT private?

What ChatGPT keeps

Everything you type or upload: prompts, files, images, voice. Plus the usual account and device data, like your name and email, your IP address, a rough location worked out from it, what you click, and the thumbs up or down you give an answer. That’s from OpenAI’s own privacy policy.

Regular chats stay in your history until you delete them. When you do, they vanish from your screen right away and are scheduled for permanent deletion from OpenAI’s systems within 30 days. You can’t get a deleted chat back.

Does ChatGPT train on your chats?

On Free, Plus and Pro, yes, by default. The setting is called “Improve the model for everyone,” and it starts switched on. Business, Enterprise and Edu accounts start with it off.

Turning it off stops future chats from being used. It doesn’t delete anything, and it doesn’t reach back for what’s already been used. And one exception most people miss: give an answer a thumbs up or down, and OpenAI says the whole conversation that goes with it may be used for training, setting or no setting.

Who can read your chats

• OpenAI, for safety

  Turning training off doesn’t stop review when a conversation may break OpenAI’s rules. Safety review is a separate thing from training.

• Contract reviewers

  In September 2026, reporting described contractors reading full conversations from personal accounts with training switched on, to rate the answers. OpenAI says names are hidden and a filter strips identifying details. It also says the filter can miss things.

• A court

  OpenAI’s privacy policy says it hands over data when the law requires it, like under a subpoena. In 2025 a court in the New York Times’ copyright case ordered OpenAI to keep consumer chats, including deleted ones, instead of erasing them. That order ended in September 2025, but it proved the logs can be held when a lawsuit wants them.

Is it private like talking to a lawyer?

No. There’s no legal privilege for a ChatGPT conversation, the way there is for what you tell a lawyer, a doctor or a therapist. OpenAI’s own CEO said so publicly in 2025, and said he thinks it should change.

Until it does, my advice is to treat a chat like an email to a company’s support line. Useful, often helpful, and not a diary.

The settings that help, and what each one really does

• Turn off training

  Settings, then Data controls, then switch off “Improve the model for everyone.” Signed in, it follows your account to every device. Or choose “Do not train on my content” at privacy.openai.com. Either one is enough.

• Use Temporary Chat

  It stays out of your history, doesn’t touch Memory and isn’t used for training. OpenAI can still keep a copy for up to 30 days for safety. Save it, and it becomes a regular chat under your normal settings.

• Check what Memory has

  ChatGPT can remember things about you across chats. Settings, then Personalization, then Memory shows what it has, and you can delete any of it. Clearing Memory doesn’t delete your chats, so do both if you want a clean slate.

• Delete it, or take it with you

  You can delete one chat, all of them, or the whole account, and you can export your data first. Requests to see, correct or delete what OpenAI holds go through privacy.openai.com.

• Leave some things out

  This one is advice, not a setting. Don’t paste in passwords, account numbers, medical records, or somebody else’s private details. Nothing above makes a chat private enough for those.