What NOT to put into AI tools (ChatGPT & co.)

AI tools (ChatGPT, Copilot, Gemini and others) are very useful at work. But they have one easily forgotten trait: everything you type into them goes to an outside company. And that is where the problems start.

Why it actually matters

• What you write may be kept and, on some services, used to “train” the tool — meaning fragments of what you put in can end up where you don’t expect.
• You lose control: once sent, you no longer know who sees it, where it ends up or how long it is kept.
• If it is client data or company confidential information, this can breach confidentiality and data protection obligations (GDPR) — and the company is liable for it.
• There have also been leaks: people have accidentally ended up seeing other people’s conversations.

In short: an AI tool is not a private box. It’s like talking to a very clever stranger who remembers everything.

The simple rule

If you wouldn’t write it on a notice board in your company’s lobby, don’t paste it into an AI tool.

What never to put in

• Client data: names, contact details, passport numbers, cards, itineraries.
• Contracts, internal documents, financial data.
• Passwords, codes, access keys.

How to use AI wisely

• If you need help with a real example, take out the real data first (replace names with “Client X”).
• Use only the tools approved by your company — some have special versions that don’t use your data.
• On any doubt, ask the IT or security team.

At home it’s the same: don’t put your own sensitive data or anyone else’s (photos of documents, ID cards, medical information) into an AI tool.