AI and Privacy: What You Should Never Paste into ChatGPT
ChatGPT is free, convenient, and remarkably chatty. But one question keeps coming up: where does everything I type actually go? Here is a plain-English look at what you should never paste into an AI, and how to stay in control of your data.
Where do your messages go?
With consumer versions, assume that what you write may be stored and, depending on your settings, used to improve the model. That is not necessarily malicious, but it is not private either. The common-sense rule: never tell an AI anything you would not tell a stranger.
What you should never paste
- Passwords, PINs, banking credentials.
- Medical data or anything deeply personal.
- Confidential company documents (contracts, client files, internal code).
- Other people's personal data without their consent (a basic GDPR reflex).
3 settings to take back control
- Turn off training.In ChatGPT's settings (Data Controls), disable model improvement.
- Use temporary chat for sensitive topics: it keeps no history.
- Anonymize before pasting. Replace real names, amounts, and addresses with [brackets].
The workplace case
At work, the free version is a bad idea for business data: no data processing agreement, data sometimes retained, unclear terms. “Pro” and “team” plans exist, with real guarantees. The real risk is unmanaged use: we cover it in EU AI Act: what your company needs to do.
Protecting your data is a skill
Understanding where your information goes is already half the battle. And it ties directly into AI-powered scams: the less data you leave lying around, the less can be used against you.
To build the right reflexes (for you and your family), start with the free quiz, then our online courses.