7 Things You Should Never Share With ChatGPT

Use AI to help you think things through while you keep control of your private information and important decisions.

ChatGPT can help you explain a confusing term, organise your ideas, and prepare questions for an expert. However, a useful conversation does not require you to share everything. Passwords, financial credentials, and identification documents should remain private, and serious health, legal, and personal decisions require human judgement.

A simple rule to follow is this: if you wouldn't post the information publicly or send it to a stranger, don't add it to an AI chat. You can remove names and identifying details, describe the situation in general terms, and still receive helpful responses. Here are seven things to avoid discussing with AI.

1. Passwords, PINs and one-time codes

Never paste passwords, PINs, one-time passwords (OTPs), account recovery codes, private keys or authenticator backup codes into ChatGPT. These details can be used to unlock accounts or approve actions. Verification codes are meant to be entered by you on the official service when signing in; no AI assistant needs to see them.

If you are locked out of your email account, for example, ask for general account recovery steps. Never send your password or the code that has just arrived by SMS. Use a unique password stored in a password manager, and enable multi-factor authentication wherever possible.

2. Bank, card and mobile money details.

Keep card and CVV numbers, online banking logins, mobile money PINs, transaction approval codes and full account details private. You do not need to reveal any financial credentials to ask for budgeting ideas or to understand a statement.

If you would like assistance with organising your monthly expenses, please ensure that you remove your name, account numbers, phone number, transaction references and any other information that could be used to identify you. Instead, you can describe the amounts by category. Contact your bank or mobile money provider via their official channels for account-specific assistance.

3. Identity documents and details that could be used to identify you.

Do not upload documents such as passports, national identity cards, driving licences, birth certificates or other documents containing your full name, ID number, address, date of birth, signature or photo. A combination of ordinary details can also identify you.

If you need help understanding a form, only share the relevant wording and remove names, numbers, QR codes, barcodes and other identifiers. Check the redacted copy carefully, as even a visible corner or document number may reveal private information.

4. Serious medical decisions.

You can ask AI to explain a medical term in plain English or help you prepare questions for a clinician. However, do not rely on AI alone to diagnose you, interpret symptoms, change medication, or decide whether to start or stop treatment. AI can make mistakes and may not be aware of your full medical history.

For example, ask “What questions should I ask my doctor about this type of test?” rather than asking AI to interpret your results. Never share your full name, hospital number, contact details or identifiable medical records. For urgent symptoms, seek qualified medical attention promptly.

5. High-stakes legal decisions.

AI can help by explaining common legal terms or by providing a list of questions to ask a lawyer. However, it may miss important facts, use outdated information, or fail to account for the law in your jurisdiction. Never let a chatbot decide whether to sign a contract, file a case, ignore a notice, or take any other step that could have serious consequences.

Instead, use it to prepare by asking for a plain-language explanation of a clause after removing names and confidential details. Then, confirm the meaning and your options with a qualified legal professional who is familiar with the relevant jurisdiction.

6. Private information about other people.

Just because you have access to someone’s messages or records does not mean you should upload them. Private chats, staff or client files, student records, health details, identity documents and confidential business information may belong to someone else or be protected by a duty of confidence.

Before using AI to summarise or analyse information, ensure that you have a valid reason and the necessary permissions. Remove names and details that could identify people, and follow your workplace, client, and service provider rules. When in doubt, describe the issue without sharing the original document.

7. Decisions you should not hand over to AI.

AI can help you compare options, identify trade-offs, and notice questions that you may have missed. However, it should not make important personal or business decisions for you. It cannot fully understand your values, relationships, responsibilities, or the implications of a choice in your life.

Rather than asking “Should I quit my job?”, explain the situation without sharing personal details and ask “What factors should I consider and what questions could help me decide?” Check the facts, talk to people you trust and make the final decision yourself.

Use AI as a helper, not as the final authority.

You don't need to avoid every sensitive topic. Leave out identifying or confidential details, ask for general explanations and verify important answers with the right person. A safer prompt might be: "I run a small business and need to understand the difference between these two general contract terms. Please explain them in plain English and suggest questions I should ask a lawyer."

Also, review the privacy and data controls of the AI service you use. In ChatGPT, the available controls depend on your account and workspace. Turning off model improvement or using Temporary Chat changes the way data is handled, but Temporary Chats may still be retained for safety purposes for a limited period. Do not use these controls as an opportunity to enter passwords, payment credentials or other information that you would not want stored or reviewed.

The safest approach is simple: share only what is necessary, remove any identifying details and use your own judgement. This allows AI to help you think more clearly without giving away control of your information or your decisions.