What Information Should You Never Put Into ChatGPT?

ChatGPT can often give you a better answer when you provide useful context.

But that doesn't mean you should tell it everything.

Just as you wouldn't enter sensitive information into an unfamiliar online form without thinking about where it was going, it's worth developing some simple privacy habits when using AI.

Here are the types of information you should think carefully about before putting into ChatGPT.

1. Passwords and login details

Never paste passwords or authentication credentials into ChatGPT.

That includes:

• passwords
• PINs
• security answers
• authentication codes
• recovery codes
• password-manager master passwords
• private access keys

ChatGPT doesn't need your actual password to help with a login problem.

Instead of saying:

"My password is [password]. Why isn't it working?"

say:

"I'm having trouble signing into an account even though I believe my password is correct. What should I check?"

You can get useful help without revealing the credential itself.

2. Banking and payment information

Avoid entering complete financial credentials such as:

• full card numbers
• bank account details
• online banking passwords
• card security codes
• PINs

If you want help understanding a transaction or statement, remove identifying information that isn't necessary for the question.

A useful question to ask yourself is:

Does ChatGPT actually need this information to help me?

Very often, the answer is no.

3. Identity documents

Think carefully before uploading documents containing information that could be used to identify you.

That might include:

• passports
• driving licences
• identity cards
• documents containing identification numbers
• official documents containing several pieces of personal information

If you only need help understanding a paragraph, consider providing the relevant wording rather than the entire document.

4. Confidential work information

This is particularly important.

Just because ChatGPT can summarise a document doesn't mean you're automatically permitted to upload it.

Your employer or organisation may have rules governing the use of external AI services.

Be especially careful with:

• confidential internal documents
• client information
• patient information
• employee records
• commercially sensitive information
• unreleased financial information
• contracts
• private meeting notes
• confidential correspondence

Check your organisation's AI, information-governance and confidentiality policies before using workplace information with an AI service.

5. Other people's personal information

Privacy isn't only about your own data.

Think carefully before entering somebody else's personal information.

That might include their:

• full name
• address
• telephone number
• email address
• financial information
• identification details
• private correspondence
• personal circumstances

Often you can anonymise the situation.

Instead of providing someone's full identity, you might say:

"A customer sent me the following complaint. Help me draft a professional response."

Then remove names, account numbers, contact details and other unnecessary identifiers from the text.

6. Sensitive health information

ChatGPT can help explain general health terminology or help you prepare questions for a healthcare professional.

But think carefully before uploading medical records or providing highly sensitive health information, particularly when the information identifies you or another person.

For a general question, identifying information may not be necessary at all.

And an AI assistant shouldn't be treated as a replacement for appropriate professional medical assessment, diagnosis or treatment.

7. Confidential legal information

The same principle applies to legal matters.

You may be able to ask a general question without providing names, addresses, case numbers or confidential documents.

If something is legally sensitive, consider whether an AI chatbot is an appropriate place for the information at all.

And don't rely on an AI-generated response as a substitute for professional legal advice where professional advice is needed.

8. Information ChatGPT simply doesn't need

This may be the most useful rule.

Before pressing send, ask:

Does ChatGPT need this information to answer my question?

Suppose you want help improving an email.

ChatGPT may need the wording of the email.

It probably doesn't need:

• your home address
• your date of birth
• an account number in your signature
• somebody else's phone number
• unrelated information elsewhere in the email chain

Remove information that isn't relevant.

This principle is sometimes called data minimisation: only providing the information necessary for the task.

You don't need to become a privacy expert to practise it.

Anonymise information where possible

Imagine you have:

"Sarah Williams at Example Company emailed me about account 12345678 and said…"

ChatGPT probably doesn't need Sarah's name, company or account number to help you respond.

You could change it to:

"A customer emailed me about their account and said…"

The problem remains understandable.

The unnecessary identifying information disappears.

This is a small habit that can make a meaningful difference.

What about ChatGPT's privacy controls?

ChatGPT provides controls over aspects of how your conversations and information are handled.

For personal accounts, you can control whether new conversations are used to help improve OpenAI's models through Data Controls.

ChatGPT also has controls relating to Memory.

Temporary Chat provides another option for conversations you don't want appearing in normal chat history or creating memories, and Temporary Chats aren't used to train OpenAI's models.

These controls are useful.

But privacy settings shouldn't replace sensible judgement about what you share.

A simple three-question check

Before giving an AI assistant information, ask:

1. Does it actually need this?

If not, remove it.

2. Can I anonymise it?

Replace names, account numbers and other identifying information with generic descriptions where possible.

3. Would I be concerned if this information were retained or seen by someone else?

If the answer is yes, think carefully about whether you should provide it at all.

That small pause can become one of your most useful AI habits.

You don't need to be frightened of AI

Privacy advice can sometimes make technology sound more alarming than it needs to be.

The answer isn't necessarily to avoid AI altogether.

It's to use it deliberately.

You already make similar decisions elsewhere online.

You wouldn't knowingly publish your password.

You wouldn't casually send your bank details to a stranger.

You wouldn't upload a confidential workplace document to an unfamiliar service without considering whether you were allowed to.

AI deserves the same kind of judgement.

The habit worth remembering

You don't have to choose between telling ChatGPT everything and telling it nothing.

Give it the context it needs.

Remove the information it doesn't.

Anonymise information where appropriate.

Understand the privacy controls available to you.

And when information is particularly sensitive or confidential, consider whether it should be entered into an AI system at all.

That's a much more useful approach than either blindly trusting AI or being afraid to use it.

Continue learning

Read Can You Trust ChatGPT? What to Check Before You Rely on an Answer to learn more about when AI-generated information needs checking.

If you're new to AI, start with ChatGPT for Beginners: A Simple Guide to Getting Started.

You can also read Is ChatGPT Free? What You Can Do Without Paying if you're exploring ChatGPT for the first time.

For a structured introduction to using ChatGPT confidently and sensibly, explore ChatGPT Made Simple.

And if you'd like practical starting points for everyday tasks, The Everyday ChatGPT Prompt Library contains prompts you can adapt rather than starting with a blank box.