Treat anything you type into an AI chatbot as something that may be stored and, in some cases, read by a person. Keep passwords, ID numbers, bank details and other people’s private information out of your chats. Then check a few privacy settings, which take only minutes to change in ChatGPT, Gemini and Copilot.
Table of Contents6 sections
This guide covers what to leave out, what happens to your chats, and where the main privacy controls sit in each app.
What Happens to What You Type?
Your chats are sent to the company’s servers, where they’re processed and usually saved to your account history. Depending on your settings, they may also be used to improve the company’s AI models.
Google’s Gemini privacy page is unusually direct about this. It says a subset of chats is reviewed by human reviewers, including trained service providers, to improve Google’s services. Google says those chats are disconnected from your account first, and that reviewed chats are kept for up to three years. Its advice is plain: don’t enter confidential information you wouldn’t want a reviewer to see.
Other providers have their own rules, which they explain in their privacy and help pages. The safest habit is the same everywhere: share only what the task truly needs.
What Not to Share With an AI Chatbot
Leave out anything that could be used to get into your accounts, steal your identity or embarrass someone if it leaked. Security companies such as Bitdefender and Norton give very similar lists:
- Passwords, PINs and verification codes. No chatbot needs them, and a real company won’t ask for them in a chat.
- ID numbers. Passport, national ID, social security and driving licence numbers.
- Financial details. Card numbers, bank account numbers and login details for banking apps.
- Your full address and date of birth together. On their own they are low risk; combined, they help with identity fraud.
- Other people’s private information. A friend’s medical news or a colleague’s contact details aren’t yours to share.
- Confidential work material. Client data, internal documents and unreleased plans. Many employers have rules about this, so check yours first.
- Private health details. If you want help understanding a topic, ask in general terms instead of pasting your records.
Strip details before you paste
You can still get help with a real document if you remove the sensitive parts first. Replace names with “Person A,” swap account numbers for “XXXX” and delete addresses. The chatbot can rewrite or summarise the text just as well.
Privacy Settings to Check in ChatGPT
ChatGPT keeps its main privacy controls under Settings, then Data controls. OpenAI’s help pages describe these options:
- Improve the model for everyone. Turn this off and your new conversations won’t be used to train OpenAI’s models. Your chats still appear in your history.
- Temporary chat. These chats don’t appear in your history, don’t create memories and aren’t used to improve the models. OpenAI says they may be kept for up to 30 days for safety purposes.
- Export and delete. Signed-in users on personal plans can request a copy of their data or delete their account from settings.
OpenAI also notes that memory is a separate setting from model training. Turning off training doesn’t clear what ChatGPT remembers about you, so review saved memories in the personalisation settings too.
Privacy Settings to Check in Gemini
Gemini’s main control is a setting called Keep Activity, which decides whether your chats are saved as Gemini Apps activity in your Google Account. Google renamed it from Gemini Apps Activity in 2025, so some menus and older guides may still use that name. Google’s privacy page explains what changes:
| Setting | What Google says happens |
|---|---|
| — | — |
| Keep Activity on | Chats are saved to your activity and may be used to improve Google’s services, including AI training, with help from human reviewers |
| Keep Activity off | Future chats aren’t used to improve Google’s AI models unless you send feedback; they’re kept with your account for 72 hours |
| Temporary chat | Not used to train Google’s AI models; also kept for 72 hours |
If you leave Keep Activity on, Google says your activity is deleted automatically after 18 months by default. You can change that to 3 or 36 months, or delete activity yourself at any time.
Feedback is shared too
Pressing thumbs up or thumbs down sends more than a rating. Google says feedback includes context such as the last 24 hours of your chat, and that it is reviewed by trained teams. Skip feedback on chats that contain anything personal.
Privacy Settings to Check in Microsoft Copilot
Microsoft Copilot has a Privacy section in its settings with switches that control whether your conversations are used to train Microsoft’s AI models. There are separate options for text and voice conversations, and the exact labels have changed over time, so look for wording about training on your conversations.
Microsoft notes that opting out of training doesn’t stop all use of your data. Its privacy page says conversations can still be used for things like safety, security and other product improvements. Work or school accounts follow their organisation’s rules instead.
Do a quick settings check whenever an app has a big update. Privacy options in AI apps change often, and new features can come with new settings.
Good Habits That Work in Any Chatbot
Use a strong password with two-step verification, delete chats you no longer need, connect only the apps you’ll use, download only official apps and never give a chatbot money or codes; these habits protect you whichever AI tool you use, including ones built into other apps.
- Use a strong, unique password and two-step verification for your chatbot account, since your chat history sits behind it.
- Delete chats you no longer need, especially any where you shared more than you meant to.
- Be careful with connected apps. Linking a chatbot to your email, files or calendar lets it read that data. Connect only what you’ll actually use.
- Watch for fake chatbot apps. Download only from the official app store listing of the company that makes the chatbot.
- Don’t trust a chatbot that asks for money or codes. Official assistants don’t need your bank login or a verification code.