Is it safe to paste my data into an AI chatbot?

Where does what I type actually go?

When you type into a chatbot, your words go to the company's servers, get processed there, and are usually saved to your account so you can scroll back later. That part is no surprise. The part people miss is the second use: on personal accounts, several companies may also use your conversations as training data to improve their next model.

Training means your chats can become part of the material a future AI learns from. Companies say they strip out names and account details first, but once a chat has gone into a model, deleting it later does not take it back out. Anthropic and OpenAI both say as much on their help pages.

The good news: every major chatbot gives you a way to say no, a way to chat without saving, and a way to delete what is already there. Here is where to find them.

ChatGPT (OpenAI)

Training. OpenAI's help page says that for its services for individuals, such as ChatGPT, "we may use your content to train our models." The switch is called "Improve the model for everyone". Business, Enterprise and Edu accounts are not used for training by default, according to the same page.

To turn it off on the web:

1. Open your account menu and select Settings.
2. Select Data controls.
3. Turn off Improve the model for everyone and select Done.

On the iOS or Android app, open the sidebar, tap your profile icon, go to Data controls and turn off the same switch. OpenAI also offers a "Do not train on my content" request in its Privacy Portal.

Temporary Chat. Start a new chat and select Temporary before sending your first message. These chats don't appear in your history, don't create memories and aren't used for training, but OpenAI says it "may keep a copy of a temporary chat for up to 30 days for safety purposes."

Deleting history. For one chat, open the ••• menu next to it in your history and select Delete. For everything, go to Settings, then Data controls, then Delete all chats. OpenAI says deleted chats are scheduled for permanent deletion from its systems within 30 days, unless it must keep them longer for security or legal reasons. Turning off training does not delete saved chats; those are separate steps.

Claude (Anthropic)

Training. Since Anthropic changed its consumer terms in 2025, people on the Free, Pro and Max plans choose whether their chats can be used to improve Claude. Existing users had to make that choice by 8 October 2025; new users make it at sign-up. Team, Enterprise, Education and API use fall under commercial terms and are not covered by this setting, Anthropic says.

To change it:

1. Select your name in the settings menu, then Settings (or go to claude.ai/settings/profile).
2. Select Privacy (claude.ai/settings/data-privacy-controls).
3. Under Help improve our AI models, turn the switch off.

The steps are the same in the mobile app.

How long it keeps chats. If you allow training, Anthropic says it may keep de-identified chats "for up to 5 years" in its training pipelines. If you don't, the existing 30-day period applies. Chats flagged by its safety systems can be kept longer: inputs and outputs for up to 2 years.

Incognito chats. In a new chat outside a project, click the ghost icon in the top right. Incognito chats aren't saved to your history or Claude's memory and aren't used for training, even if you allowed training. Anthropic keeps them for 30 days by default.

Deleting history. Open a conversation, click its name at the top and choose Delete. To remove several, open your Recents list, choose View all, tick the ones you want and click Delete Selected. Anthropic says deleted chats leave its back-end storage within 30 days, and a deleted chat won't be used for future training.

Gemini (Google)

Training. Google's setting is called Keep Activity, and it is on by default for users aged 18 and over. While it is on, Google says your activity is used to "provide, develop, and improve its services (including training generative AI models)." Some chats are also read by human reviewers; Google keeps those for 3 years, disconnected from your account, even if you delete your activity.

To turn it off:

1. On a computer, go to gemini.google.com, open Settings & help and choose Activity (or go to myactivity.google.com/product/gemini).
2. Near the top, click On, then choose Turn off or Turn off and delete activity.

With Keep Activity off, Google still keeps chats for up to 72 hours to run the service, but they aren't used to train its models unless you send feedback.

How long it keeps chats. By default, activity is auto-deleted after 18 months. On the same Activity page you can choose 3 or 36 months, or no auto-delete.

Temporary chat. On gemini.google.com, signed in with a personal Google account, click Temporary chat next to New chat. Google says temporary chats are not used to train its models and are kept for 72 hours.

Deleting history. On the Activity page, click Delete activity item next to a single chat, or click Delete above your activity and choose a range such as Last hour, Last day or All time.

Microsoft Copilot

Training. Microsoft's privacy FAQ for the older Copilot app says it uses Copilot conversations, including voice, for training unless you opt out or fall into an excluded group. Excluded groups include people who aren't signed in, signed-in users under 18, people using a work or school (Entra ID) account, and users in a handful of countries. Microsoft says an opt-out applies to past, present and future conversations and takes effect within 30 days. Human review of some conversations cannot be switched off.

In the older app, the steps were:

1. On copilot.com, click your profile icon, then your profile name, then Privacy. (On Windows or macOS: profile icon, Settings, Privacy. On mobile: menu, profile icon, Account, Privacy.)
2. Turn off Training on conversation activity and Training on voice conversations.

Microsoft released a new version of the Copilot app on 18 August 2026, and the help pages for that version that we checked do not describe a training switch.

How long it keeps chats. "By default, we store conversation activity for 18 months," according to the FAQ.

Temporary chat. None of the Microsoft help pages we opened mention a temporary or private chat mode.

Deleting history. In the new app, open the More menu next to a conversation in the Chats list and choose Delete. To delete your whole history, Microsoft points you to the Microsoft privacy dashboard. Saved memories are separate: go to Settings, Personalization, then Manage next to Saved memories.

Meta AI

Training. Meta's help page puts it plainly: "Meta uses your interactions with AIs to improve AI at Meta." In the EU, Meta announced in April 2025 that it would train its models on public posts and comments from adults, and on the questions people send to Meta AI. It says it does not train on private messages with friends and family, unless someone in the chat shares them with Meta AI. Public data from EU users under 18 is excluded.

On the pages we could open, Meta does not describe an in-app switch to stop your Meta AI chats being used for training. For people in the EU, the route Meta gives is an objection form, linked from its in-app and email notices, which it says you can use "at any time".

How long it keeps chats. Meta's help page does not give a retention period for Meta AI chats.

Deleting history. In the Meta AI app:

1. Tap Menu in the top left, then Settings in the bottom left.
2. Under App settings, tap Data & privacy, then Manage your information.
3. Tap Delete all chats and media, then Delete all.

Meta's page does not cover deleting Meta AI chats inside WhatsApp, Instagram or Facebook.

What should I never paste into a chatbot?

Opting out of training reduces one risk. It does not make a chatbot a safe. Your chats are still stored for a while, may be read by human reviewers, and could be exposed in a breach or a legal request. A sensible rule: don't type anything you wouldn't put in an email to a stranger at a large company.

• Passwords, PINs and security codes. No legitimate chatbot task needs them.
• ID and account numbers: passport, national ID, tax number, bank card or IBAN details.
• Other people's private information, especially health records, a friend's medical test or a colleague's personnel file. They haven't agreed to share it.
• Confidential work documents, unless your employer has given you a tool approved for that. More on this below.

If you want help with a document that contains personal details, replace the names and numbers with placeholders such as "Client A" first. The chatbot rarely needs the real ones. Our guide to writing a good prompt shows how to give context without oversharing.

Is a work account different from a personal one?

Usually, yes. OpenAI says it does not train on ChatGPT Business, Enterprise or Edu by default. Anthropic's commercial plans sit outside its consumer training setting. Microsoft excludes work and school accounts from Copilot training, and Google warns that Gemini on a work or school account "may be subject to different data handling terms." In those setups, your employer usually controls retention and can often see or export what you typed.

So the practical rule runs both ways. Use the tool your employer provides for work material, and keep personal matters off your work account. Pasting a client contract into your personal chatbot can break your employment contract even if the chatbot never trains on it. Our AI at work guide covers this in more depth.

What rights do I have in the EU?

If you live in the EU, the General Data Protection Regulation (GDPR) applies to chatbot companies just as it does to your bank. According to the EU's official Your Europe portal, you have the right to get a copy of the personal data a company holds about you, free of charge, and the company should reply within one month. You can also ask for your data to be erased when it is no longer needed or is being used unlawfully, often called "the right to be forgotten."

If a company ignores you, you can complain to your national data protection authority, which should respond within three months. The Your Europe page on data protection explains each right in plain language. Most chatbot makers also offer a privacy portal or request form for these requests, so start there.

For the wider rules on AI itself, see our guide to the EU AI Act.