An AI has to say it is an AI: the transparency the law requires

Of the whole EU AI Act, there is one obligation that directly affects any assistant that talks to clients: transparency. Put briefly, whoever converses with an artificial intelligence system has the right to know they are not talking to a person.

General guidance, not legal advice.

What it says exactly

The regulation establishes that AI systems intended to interact directly with people must be designed so those people are informed that they are interacting with an AI, unless it is obvious from the context. This obligation is part of the block that enters general application in August 2026.

It is a deliberately simple requirement, and precisely for that reason it leaves no room for tricks: hiding it in terms and conditions nobody reads is not enough.

Why this is good news

It might look like a commercial nuisance — “if it says it's a robot, people will distrust it” — but practice says otherwise.

What generates distrust is not talking to an AI: it is discovering halfway through a conversation that you thought you were talking to a person and you were not. That sense of having been misled does have a cost, and it is paid in the relationship with the client.

When it is clear from the start, something interesting happens: people adjust their expectations and use the tool better. They ask what an AI can answer — hours, availability, indicative prices, booking — and ask for a person when they want a person. The conversation works better for everyone.

Transparency, properly understood

Complying is not about firing off a legal notice before every message. It is about the nature of the other party being obvious:

1. Introduce itself naturally. Make it clear from the start of the conversation, without cold formulas.
2. Do not pretend to be a person. No invented employee names, and no pretending to be “checking with the team” when that is not what is happening.
3. Acknowledge the limits. A “my colleague can confirm that for you” is transparency, and it is also good service.
4. Always offer the human exit. Asking to speak to someone should be easy and should work.

And a specific note on voice: when the assistant takes calls, the requirement is the same. A natural voice is not a licence to pass yourself off as someone.

How ClaudIA does it

Here we have not had to change course, because that was the approach from the beginning: ClaudIA introduces itself for what it is, an AI virtual assistant. It never passes itself off as a member of the team, and when a conversation needs human judgement it alerts someone at the business and leaves them the context instead of improvising.

In the beauty sector that means not pretending to be the person who looks after you at the salon. In a dental clinic it also means never giving an opinion on a clinical case: it informs and books, and the judgement belongs to the professional.

If you want the full map of the rules, start with what the EU AI Act means for your business.

Transparency takes nothing away: it is what turns automation into something your clients appreciate rather than endure.