When a Phonica voice agent picks up, the first thing it does is introduce itself: it says your company's name and that it is a virtual assistant. That isn't a courtesy or a design choice: it has been a legal obligation across the European Union since the summer of 2026, and it is also what makes the conversation work.
What the law says
The EU Artificial Intelligence Act (Regulation (EU) 2024/1689, known as the "AI Act") devotes its Article 50 to transparency. As far as a voice agent is concerned, it says this: AI systems intended to interact directly with people must be designed so that those people know they are interacting with an AI, unless it is obvious to someone reasonably well-informed and observant, taking the context into account.
On a phone call it is not obvious. Today's voices sound natural and the caller has no way of knowing unless someone tells them. So, for a phone agent, the obligation is clear: it has to say it is an AI.
That obligation applies from 2 August 2026, the date on which most of the Regulation became applicable. And the penalties for breaching the transparency obligations are the kind no business wants to risk: the Regulation provides for fines of millions of euros or a percentage of annual turnover.
How it applies to a voice agent
In practice, compliance comes down to three things, and all three are designed into the agent itself:
- It introduces itself when it picks up. "Hello, you've reached [your company]. I'm its virtual assistant, how can I help you?". Clear, at the start and with no small print.
- It doesn't pretend to be a person. If someone asks "are you a real person?", it tells the truth. An agent designed to dodge that question breaks the rule and, besides, destroys trust the moment it's found out.
- It lets callers speak to a person. The law doesn't put it in those terms, but it is the natural consequence: someone who knows they are talking to an AI should be able to ask for a member of your team, and the agent transfers or takes a message.
And the recording notice
There is a second notice that sometimes gets confused with the first. If the call is recorded (and with a voice agent it usually is, so that you can review in your dashboard what was said), data protection law, the GDPR and its implementation in Spain, requires callers to be informed. It's the "this call may be recorded" line you already know, and it is independent of the notice that they are talking to an AI.
Our agents include both notices where applicable, at the start of the call, in a single natural sentence that scares nobody off. And the data from each call is processed only to provide the service, under a data processing agreement with your company, as we explain on the services page.
Why it's good for you too
It might seem that saying "I'm an AI" drives callers away. Our experience is the opposite, and it makes sense:
- The caller adjusts their expectations. They know they can get straight to the point, that there's no need to be diplomatic, and that the routine stuff will be resolved on the spot.
- What makes people hang up isn't the AI. It's voicemail, keypad menus and waiting. If the agent understands and resolves, people carry on talking as normal.
- Trust is won at the start. A customer who finds out mid-call that they weren't talking to a person feels deceived. One who knew from the first second doesn't.
- Your business looks good. Being honest about the technology you use says something about how you treat your customers.
A voice agent that identifies itself doesn't lose calls: it loses the caller's doubts.
What we do at Phonica
- All our agents introduce themselves as a virtual assistant at the start of the call. From day one, before the law required it.
- They include the recording notice when the call is recorded.
- They never deny being an AI and always let the caller ask for a person.
- The data from each call is kept separate per company and processed under a data processing agreement.
If you want to hear what a well-done introduction sounds like, talk to our AI: the first thing it does is tell you what it is.
Frequently asked questions
Is a voice agent required to say it is an AI?
Yes. Article 50 of the EU Artificial Intelligence Act requires that people know they are interacting with an AI system, unless it is obvious. On a phone call it is not obvious, so the agent has to say so.
Since when does it apply?
That transparency obligation applies from 2 August 2026. Phonica's agents have identified themselves from day one, before it became enforceable.
Do callers also have to be told the call is recorded?
Yes. If the call is recorded, callers must be informed at the start under data protection law (GDPR). It is a separate notice from the AI one, and our agents include both where applicable.
Do people hang up when they know they're talking to an AI?
In our experience, no: what makes people hang up is voicemail or a keypad menu. If the agent introduces itself naturally, understands what it's told and resolves the request, most people carry on the conversation as normal. And anyone who prefers a person can ask for one at any time.
Published 23 October 2026. This article explains the rules in plain terms and is not a substitute for legal advice on your specific case. Source: Regulation (EU) 2024/1689, Article 50 and Article 113 (application dates).