Skip to content
AI Digested
GuideBeginner3 min read

AI and your privacy: what happens to what you type into a chatbot

Do AI chatbots keep your conversations? Are they used for training? Who can read them? A plain guide to AI privacy, and the settings worth changing today.

First published · AI Digested

Chatting with an AI feels private, like talking to yourself. It isn't. What you type goes to a company's computers, and what happens next depends on that company's policies and your settings.

What usually happens to your chats

With a typical free or personal account on a major chatbot:

  • Your conversations are stored, so you can come back to them.
  • They may be used to train future models, unless you opt out. Training means your text could influence how future versions behave.
  • Some may be reviewed by people, for example to check for abuse or rate the AI's answers.
  • They may be kept after you delete them, for a period set out in the privacy policy, and longer if required by law or a court order.

Policies differ between companies and change often. The general pattern is the same: free consumer accounts give companies the most use of your data; business accounts give the least.

Settings worth checking today

Every major chatbot lets you control at least some of this. Look in the app's Settings, usually under Data controls or Privacy, for:

  • Training on your chats. Look for an option like "Improve the model for everyone", "Help improve" or "Activity". Turning it off stops your future chats being used for training.
  • Chat history and memory. Many assistants now remember facts about you across chats. You can usually view, edit or turn off memory.
  • Temporary or incognito chats. These aren't saved to your history, and generally aren't used for training, although they may still be kept briefly for safety.
  • Connected apps. If you've linked your email, calendar or files, review what each connection can access.

What never to type into a chatbot

  • Passwords, PINs and security codes. No real service will ever need them in a chat.
  • Full bank or card details.
  • Other people's personal information: names with health details, addresses, ID numbers.
  • Confidential work material, unless your employer has approved the tool. Samsung banned staff from using public chatbots in 2023 after engineers pasted in sensitive code.
  • Anything you'd be upset to see leaked. Breaches and bugs happen. In 2023 a bug briefly showed some ChatGPT users the titles of other people's chats.

Work and school accounts are different

Business, enterprise and education versions of ChatGPT, Claude, Gemini and Microsoft Copilot generally promise not to train on your data by default, and come with contracts and admin controls. If your employer provides one, use that for work, not your personal account.

Running AI privately

For the most privacy, you can run an open-weight model on your own computer. Nothing leaves your machine. These models are less capable than the biggest cloud chatbots, but good enough for many tasks. How to run AI on your own computer.

In the UK and EU, data protection law (GDPR) gives you the right to know what personal data a company holds about you, to have it corrected or deleted in many cases, and to object to some uses. Italy's data protection authority temporarily blocked ChatGPT in 2023 over privacy concerns, and European regulators continue to scrutinise AI companies.

Keep going

Common questions

Can AI companies read my chats?

Usually, in limited circumstances. Companies typically keep conversations for a period, and staff or contractors may review some, for example to investigate abuse or improve the model. Each company's privacy policy explains when.

Does deleting a chat delete it completely?

Not always immediately. Deleted chats are usually removed from your history straight away but may be kept on the company's systems for a period, often around 30 days, and longer if needed for legal reasons or safety investigations.

Is it safe to use AI for health questions?

It can help you understand terms and prepare questions for a doctor. Be thoughtful about how much identifying detail you include, and never treat the answer as a diagnosis.

What about AI in apps like WhatsApp or Gmail?

When you use an AI feature inside an app, the content you share with that feature is processed by the AI. Check the app's settings and privacy information for what's stored and whether it's used for training.

Spotted a mistake or something out of date? Tell us and we will fix it. Read how we research and write our articles.