6 min read
Is ChatGPT safe to use?
Short answer: yes, for everyday things, as long as you treat it like a very capable stranger rather than a private diary. Here is exactly what to change, what to avoid, and how to check what it tells you.
Start here
- Turn off 'improve the model for everyone' in Settings → Data controls (2 minutes).
- Swap real names and numbers for placeholders like [CLIENT NAME] before pasting.
- Ask 'what in this answer might be wrong or out of date?' before you act on it.
What actually happens to what you type
Your words travel over the internet to the company running the tool, are answered by their computers, and — depending on one setting — may be kept and used to help train future versions.
Nobody is sitting reading your chats for fun. But 'stored on someone else's computer' is the right mental model, the same as an email you send.
- Chat history: on by default, so you can find old conversations.
- Training on your chats: usually on by default, and switchable off.
- Temporary chat: a one-off conversation that is not saved at all.
The five-minute privacy setup
Do this once and you can relax afterwards. In ChatGPT, click your name or picture at the bottom left, then Settings, then Data controls.
- Turn off 'Improve the model for everyone'.
- Use 'Temporary chat' for anything sensitive — it disappears afterwards.
- Delete old conversations you would not want read aloud.
- Use a strong, unique password and turn on two-step login.
Things to never paste in
Treat these as off limits, in any AI tool, no matter how convenient it feels in the moment.
- Bank details, card numbers, passwords or PINs.
- Passport, national insurance or ID numbers.
- Someone else's medical or financial information.
- Work documents covered by a confidentiality agreement.
Safe way to get help with a real document
Here is a letter with the personal details replaced by placeholders like [NAME] and [ACCOUNT NUMBER]. Please rewrite it so it is polite, clear and under 200 words, and keep the placeholders exactly as they are.
The bigger risk: it can be confidently wrong
AI writes fluently even when it is mistaken, and it never sounds unsure. For most people that is the real danger — not privacy, but believing something untrue because it was well written.
It is especially shaky on prices, dates, opening hours, law, medical advice, and anything that changed recently.
Paste this after any answer that matters
Which parts of that answer are you least confident about, and which parts should I verify with an official source? List them as bullet points, shortest first.
How to check something in under a minute
You do not need to be technical to fact-check. You need one habit: never let AI be your only source for anything with money, health or a deadline attached.
- Search the key claim in your normal search engine and look for an official site.
- For anything about a company, check that company's own page.
- For health or legal matters, treat AI as a way to prepare questions, not answers.
Common mistakes
- Pasting a full bank statement to 'help work out my spending'.
- Trusting a phone number, price or opening time it invented.
- Assuming a private-looking chat window means a private conversation.
Questions people ask us
Can other people see my chats?
Not other users, no. Your conversations are not shared publicly or shown to anyone else using the tool.
Is the free version less safe than the paid one?
The safety settings are the same. Paid mostly buys you speed and better models, not extra privacy.
Should I use my real name?
It is fine for the account. Inside conversations, there is rarely a reason to include full names of other people.
Change one setting, keep private numbers out, and fact-check anything with money or health attached. Then use it freely.