AI at work: the confidentiality habits

An AI assistant is a formidable accelerator. But at the office, every text you paste into it leaves your machine. Before making it a habit, adopt these five precautions.

1. Assume everything you paste can be processed elsewhere

Depending on the tool and the plan, your inputs may pass through third-party servers, or even serve to improve the service. That is not necessarily a problem, but it is the starting point: treat every input as data that leaves your premises.

2. Never paste sensitive data without a guarantee

Clients’ personal data, health information, trade secrets, credentials: by default, they do not go into a consumer assistant. This is especially true for meeting transcription tools, which capture exchanges that are often sensitive. If you need to, you need a professional plan with clear commitments on data handling.

3. Distinguish the free plan from the professional plan

Professional versions generally offer guarantees absent from free versions: no use of your data for training, controlled hosting, access management. For use at work, that is often the minimum condition.

4. Anonymise where possible

You can often get useful help without handing over the real data. Replace names, amounts and identifiers with fictional examples. The model reasons just as well on an anonymised case.

5. Check your organisation’s stance

Many companies have a policy on AI use, sometimes a list of approved tools. Before scaling up a use, find out. In France, the data protection authority (CNIL) publishes useful markers to frame these practices, and depending on your sector, the European AI Act already sets obligations.

Key takeaway

AI at work is not forbidden, it is to be framed. One principle is enough to avoid most slip-ups: only entrust an assistant with what you would accept leaving your organisation. For the rest, anonymise or use a professional plan.