AI Tool Privacy Guide: ChatGPT, Gemini, Copilot & Claude

The Short Answer: Which AI Tools Are Private?

None of the major AI tools are truly private, but they vary in how easy they make it to improve your privacy.

ChatGPT: Stores conversations and uses them for training by default. Opt-out available in Settings → Data Controls (takes 2 minutes).

Google Gemini: Stores conversations and may use for model improvement by default. Opt-out available in Settings (takes 2 to 3 minutes).

Microsoft Copilot: Data handling depends on your account type (consumer vs work). Opt-out available in Settings (takes 2 minutes).

Claude (Anthropic): Best default privacy of the four because conversations are not used for training by default (though they are stored).

Never type passwords, bank details, or NHS numbers into any tool. Read on for tool-specific opt-out steps and a comparison table showing which tool has the best default privacy.

Privacy Settings Comparison: Which Tool Has the Best Privacy?

All four major tools store conversations. The differences are in data retention, training use, and how easy they make it to opt out.

Tool	Training Use	Data Retention	Opt-Out Available	Difficulty
ChatGPT	Yes, by default (free/Plus)	30 days minimum	Yes (Settings → Data Controls)	Easy (2 min)
Gemini	Yes, by default	3 months	Yes (Settings → Activity controls)	Moderate (3 min)
Copilot	Depends on account type	Varies	Yes (Settings)	Moderate (2 to 3 min)
Claude	No, by default	30 days	Not needed (no training use)	N/A

Takeaway: Claude has the best default privacy (no training use). ChatGPT and Gemini require opt-out to prevent training use. All four store conversation history for service purposes.

What Does OpenAI Do with Your Conversations?

When you use ChatGPT, OpenAI receives and stores the full conversation. That includes everything you type and everything ChatGPT writes back.

By default, on a free or Plus account, OpenAI may use those conversations to improve their AI models. They also retain data for at least 30 days after you delete a chat, for safety monitoring purposes.

There is a difference between ChatGPT and a search engine that is worth understanding. When you search Google, Google records what you searched for. When you use ChatGPT, OpenAI records the full conversation, including everything you wrote. That is a more detailed record.

OpenAI staff can access conversations in certain circumstances, such as safety investigations. Their privacy policy sets this out in full. I read it so you do not have to. The part that matters most for everyday users: your conversations are stored, and OpenAI may use them for training unless you opt out.

How to Stop Your Chats Being Used for AI Training

You can opt out by changing one setting. Here is how to find it.

Step 1. Log into ChatGPT at chat.openai.com.

Step 2. Click your name or profile in the bottom-left corner. On some browsers it appears as a small circle with your initial.

Step 3. Click "Settings", then "Data controls".

Step 4. Find the toggle labelled "Improve the model for everyone" and switch it off.

Once off, your new conversations will not be used for training. Conversations you had before the change may already have been processed.

If you use a Team or Enterprise account through work, your data is excluded from training by default. Nothing to change.

Google Gemini Privacy Settings & Opt-Out

Gemini stores your conversation history and may use it to improve Google's AI models by default.

Step 1. Go to myactivity.google.com or access Settings within Gemini on your device.

Step 2. Click "Manage all Web and App Activity", then toggle off "Web and App Activity" to stop Google from using your Gemini conversations for model training.

Alternatively (in Gemini app directly): Open Gemini, click Settings (top-right), then "Data and Privacy", then toggle off "Improve Gemini for everyone".

Google retains conversation data for approximately 3 months even after deletion, for security review. If you want faster deletion, use Gemini's temporary chat feature (look for the pencil icon and select "Temporary chat").

Microsoft Copilot Privacy Settings & Opt-Out

Copilot's privacy settings depend on whether you are using a consumer account or a work/school account.

For consumer accounts:

Step 1. Open Copilot at copilot.microsoft.com and click your profile icon (top-right).

Step 2. Go to Settings → Privacy & Cookies and review the toggle for "Help improve Copilot". Turning this off prevents your conversations from being used for model improvement.

For work/school accounts: Your organization controls privacy settings. Check with your IT team; most enterprise accounts have training use disabled by default.

Microsoft stores conversation logs for approximately 30 days. Use Copilot's private chat mode (if available in your region) for more sensitive queries.

Claude Privacy Settings & Opt-Out

Claude, made by Anthropic, has the most user-friendly privacy defaults of all four tools. Conversations are not used to train Claude by default.

The simple truth: You do not need to opt out of Claude to prevent training use. It does not happen automatically. Conversations are stored on Anthropic's servers for 30 days (for safety and billing), then deleted.

If you want additional control:

Step 1. Log into claude.ai and go to Settings (bottom-left).

Step 2. You will see "Chat History" is on by default. Turn this off if you want conversations deleted immediately after each session, rather than stored for 30 days.

Turning off chat history means you lose access to past conversations, but also means no log exists on Anthropic's servers beyond the current session. This is useful if you are testing Claude with sensitive information and want immediate deletion.

Anthropic's transparency reports (published annually) show they do not share user data with third parties and do not use it for advertising. Their privacy page is clear and worth reading.

Is It Safe to Type Confidential Information Into Any AI Tool?

No. Do not type confidential information into ChatGPT, Gemini, Copilot, or Claude.

Confidential means anything you would not post publicly: passwords, bank account numbers, National Insurance numbers, NHS record details, legal correspondence, work documents with client names, photographs of identity documents. If you are unsure whether something counts, assume it does.

Conversations in all four tools are stored on their respective company servers, not just on your device. They are not end-to-end encrypted the way a messaging app like Signal is. And while each company has security in place, no online service is entirely proof against data breaches or internal access.

None of this makes these tools dangerous for everyday use. General questions, drafting help, public information lookups, and creative tasks carry very little risk. The key is keeping personal and confidential details out entirely. For a full list of what to keep out, see our guide on what not to tell ChatGPT.

Is ChatGPT Private? What That Actually Means

"Safe" and "private" are different things, and the difference matters here. Safe means the service has security measures to protect your data; ChatGPT meets a reasonable standard on this front. Private means only you can see the content. ChatGPT does not meet this standard. Your conversations are stored on OpenAI's servers, and in certain circumstances staff can access them.

Think of it less like a sealed letter and more like writing on a postcard. It gets where it needs to go, and it is generally protected. But it is not sealed from everyone at the company.

ChatGPT is not end-to-end encrypted the way WhatsApp or Signal messages are. End-to-end encryption means only you and the other party can read the content, not even the company running the service. ChatGPT does not work that way.

For most everyday conversations, this barely matters. Asking for recipe ideas, help writing a card, or a film recommendation carries very little risk. The concern arises when people type personal details, medical records, confidential work files, or anything with names and addresses. Those things should stay out. Our complete ChatGPT safety guide covers the broader picture.

Your UK Rights Under GDPR

If you live in the UK, you have rights under the UK GDPR (General Data Protection Regulation) that apply to how all four AI tools handle your data.

Your main rights include:

• The right to access. You can ask for a copy of the data each company holds on you.
• The right to deletion. You can ask them to delete your personal data, subject to certain legal retention requirements.
• The right to object. You can object to your data being used for AI training. Opt-out toggles in each tool's settings are the quickest way to act on this.
• The right to data portability. You can ask for your data in a machine-readable format.

Contact details for formal data requests:

• OpenAI: privacy.openai.com or email dsar@openai.com
• Google (Gemini): Visit your Google account settings or submit a request via Google's Data Download & Privacy page
• Microsoft (Copilot): Visit your Microsoft privacy dashboard or contact their privacy team
• Anthropic (Claude): Email privacy@anthropic.com or visit anthropic.com/claude-privacy

The ICO (Information Commissioner's Office) has published guidance on AI tools and your rights when using AI services in the UK. Their guide to getting copies of your information explains how the data access request process works.

Companies must respond to formal requests within 30 days under UK GDPR law.

What IS Safe to Type into ChatGPT?

Much coverage of ChatGPT privacy focuses on what to avoid. That is useful, but it misses something: most things people want to ask are completely fine.

You can safely type:

• General questions about history, science, how things work, or current affairs
• Recipe ideas and cooking questions
• Help drafting a letter or email (without real names, addresses, or sensitive context)
• Film, book, TV, or podcast recommendations
• Grammar checking or writing feedback on draft text
• Ideas for a day out or holiday (general suggestions, not your home address)
• Creative writing or story ideas
• General medical topics, as distinct from your specific records or symptoms

I use ChatGPT daily for drafting and research. I keep medical questions, anything with names and addresses, and work documents with client information entirely off it. The line I draw: if I would be comfortable with a support technician at a large company potentially seeing it, it goes in. If not, it does not.

If you also use other AI tools, our guide to Is Meta AI safe? covers the version now built into WhatsApp.

How to Use Temporary Chat for Sensitive Topics

Temporary Chat is a feature that lets you have a conversation with ChatGPT that is not saved to your history and, according to OpenAI, will not be used to train their AI.

We checked: when you use Temporary Chat, the conversation does not appear in your sidebar history after the session ends.

To start one, look for the pencil or "new chat" icon at the top of the page. You will usually see a "Temporary chat" option there. The exact location shifts with updates, but the feature has been available since early 2024.

Keep in mind that OpenAI may retain brief logs even from temporary sessions for safety monitoring. But those conversations will not sit in your account history or be used for model training.

How to Delete Your Data Across All AI Tools

All four tools let you delete your conversation history. The process differs slightly for each.

ChatGPT: Go to Settings → Data Controls → Manage → Delete all chats. Confirm. OpenAI will retain metadata for 30 days, then delete everything.

Gemini: Go to myactivity.google.com or Settings → Data & Privacy → Delete Gemini activity. Choose your date range and delete. Google retains some logs for 3 months for safety review.

Copilot: Go to Settings → History and activity → Clear activity. Note: clearing history does not delete Microsoft's safety logs; those remain for compliance purposes.

Claude: Go to Settings → Chat History → turn off to delete immediately. Or delete individual chats by clicking the trash icon next to each conversation. Anthropic deletes chats 30 days after you delete them from your account.

Formal data deletion request: Under UK GDPR, you have the right to request deletion of all your personal data. Contact each company's privacy team (email addresses in the GDPR section below). Expect a response within 30 days.

For instructions on deleting chats you have already had, see our guide on how to delete your ChatGPT history.

FAQ: AI Tool Privacy Questions Answered

• Complete ChatGPT safety guide A broad look at safety, risks, and good habits for new users.
• What not to tell ChatGPT A practical list of things to keep out of all AI tools.
• Staying safe with AI Our main hub for AI safety guidance.
• OpenAI Privacy Policy What OpenAI stores and how they use your data.
• Anthropic Claude Privacy Anthropic's privacy page for Claude users.
• ICO guidance on AI tools The UK's data regulator explains your rights.

Want to learn how to use all these AI tools confidently and safely? The WellWired AI Academy covers ChatGPT, Gemini, Claude, privacy, and everyday AI at your own pace. Thirty modules, designed for beginners.