Is ChatGPT safe for personal information? What to share, what to skip

Almost everyone asks the same thing within a week of trying an AI chat tool: is it okay to type this in? A résumé, a letter from the bank, a message about a family problem. It's a sensible question, and the honest answer is more useful than "yes" or "no".

Here's the short version. You can't see exactly what happens to a conversation after you send it, and the rules differ between companies and change over time. So the safest habit isn't to memorize policies. It's to share less. In most cases the AI can help you just as well without your private details, and that's easy to arrange.

Bad vs. good: the same request, two ways

A simple rule of thumb

Before you paste anything, ask: "Would I be comfortable if a stranger read this?" If not, remove the sensitive part or swap it for a placeholder. That one question catches most problems.

Things most people are wise to keep out of a chat:

• Account and ID numbers: Social Security number, passport or driver's license numbers, bank and card numbers.
• Passwords and codes: never type these into a chat, for any reason.
• Full identifying details: your home address, phone number, birthdate, and other people's too.
• Other people's private information: a friend's health news, a coworker's complaint, a child's full name and school.
• Confidential work material: customer lists, unreleased plans, anything your employer asks you to keep inside the company.

Things that are usually fine: general questions, drafts with names removed, topics you're learning about, and anything you'd happily post on a public bulletin board.

The placeholder trick

This is the single most useful habit. Instead of real details, type a label in square brackets and fill in the real thing later, after the AI is done.

You get a ready-to-send draft, and the only copy of the sensitive details is the one on your screen. It works for letters, emails, forms, and complaints. For work messages specifically, our work email prompts pair well with this.

Clean up a document before you paste it

Sometimes you do want help with a real document, like a résumé or a long email thread. Let the AI help you remove the sensitive parts first, or do it yourself in a minute.

For résumés, removing your home address, phone, and email before pasting is a good default; the AI doesn't need them to improve your wording. Our résumé prompts guide shows the full process.

Ask general questions, not personal ones

Many sensitive topics, like health, money, and legal questions, can be learned about in general terms without putting your own details in. You describe the situation, not the person.

This is general education only, not advice for your situation. For medical, legal, or financial decisions, a qualified professional who knows your facts is the right person, and a good use of the AI is helping you prepare questions for them.

Describe a situation without identifying anyone

Check your settings and the privacy policy

Placeholders are your first line of defense. Your second is knowing what your tool does with your conversations. The details vary and change, so check the current version rather than relying on a blog post, including this one. Look for:

• Chat history and training: many tools have a setting about whether your conversations can be used to improve their models. It's usually in Settings under something like "Data controls" or "Privacy".
• Temporary or incognito chats: some tools offer a mode that doesn't save the conversation to your history. Handy for one-off questions.
• Deleting conversations: find out how to delete a chat and what that does.
• Work or school accounts: if your employer or school provides an AI tool, it may have different rules. Check them before pasting anything work-related.

That last line matters. An AI can be wrong about its own company's current policies, so treat its answer as a reminder of what to look for, not as the final word.

Why this works

A prompt is just text you give the AI to work with, and the AI can only use what you provide. Brackets and general descriptions give it the shape of your situation, which is usually all it needs. If you're new to this idea, what a prompt is is a good starting point. You're not making the answer worse by leaving details out; you're keeping control of the part only you should hold.

ChatGPT vs Claude for privacy

Both ChatGPT and Claude are widely used, and both have privacy settings and published policies. We can't promise how either handles your data today, because the terms differ by plan (free, paid, work) and companies change them. In practice the same habits apply to both: share less, use placeholders, look at the data settings once, and read the current policy for your plan. If you're choosing between them for other reasons, see our ChatGPT vs Claude comparison for beginners.

A few honest cautions

Deleting a chat isn't a guarantee of anything. Treat what you type as something that might be stored somewhere for a while.

Don't paste other people's secrets. It's not just about your own information. A friend who told you something in confidence hasn't agreed to have it typed into a tool.

Don't trust AI answers about your own accounts. Never enter a password, and be wary of any chat or website that asks you for one.

Want more ready-to-use prompts for work, home, and everyday tasks? Get 50 prompts like this free (PDF).

Frequently asked questions