At work, what you can paste into an AI assistant, and what you keep

The question comes up in every training session: “can I put this into the AI?”. The answer depends less on the assistant than on where your data goes and what your company has set up. Here are the five families to keep for yourself, and how to work anyway.

The method in five points

Personal data

Names, contact details, health data, appraisals of clients or employees. The GDPR applies, and the consumer tool is often not covered by a contract with your company.

Non-public figures

Results, margins, budgets, forecasts. They are worth money, sometimes even more to a competitor.

Trade secrets

Processes, formulas, source code, unannounced projects, price lists. A leak cannot be undone.

Confidential contracts and files

Contracts under negotiation, litigation, HR files. Their confidentiality is often a contractual obligation.

Credentials

Passwords, access keys, card numbers. Never, anywhere, not even to “check a syntax”.

What AI does not do for you

In most cases the assistant does not need the sensitive data to help you: “an important client” is enough, their name adds nothing. Replace names with markers, round the amounts, remove credentials. And if your company has no approved assistant yet, the risk is real: that is shadow AI.

What stays out of the request

A company can set up an assistant whose data stays in-house, with a contract that forbids training on its data. It is then the right place for sensitive uses, and the way to make the rule simple for everyone.

Find out where you stand

Before choosing a course, the most useful step is knowing where you start. Our AI skills test asks questions matched to your role and tells you, in a few minutes, what you already do well and what would save you the most time. For a whole team, AI training for teams starts from your real tasks, not from a generic course.