Does an AI phone agent have to say it's an AI?
By Alperen Zeki Gokmen, AI Builder & Educator
Updated 3 min read
The short answer
Yes, in the EU, since August 2, 2026. Article 50 of the AI Act says people must know they are talking to an AI from the first moment. The Commission's guidelines ask phone agents to say it out loud at the start of the call. An agent that books visits should also say who it works for.
For clinics, studios and coaches in the EU who want an agent to pick up calls after hours.
What does the law say?
Article 50(1) of the EU AI Act covers AI systems that talk directly with people. They must be built so "the natural persons concerned are informed that they are interacting with an AI system". For a business, the exception that matters is when that is already obvious to a reasonably well-informed, observant person.
Article 50(5) sets the timing. The information comes "at the latest at the time of the first interaction". The European Commission says the rule applies from August 2, 2026.
How should a phone agent say it?
The Commission's guidelines of July 20, 2026, give examples. For voice and phone, they ask for "explicit spoken statements at the beginning of the interaction". In longer calls they suggest reminders, especially after an interruption. A tone or beep on its own isn't enough.
Agents that make bookings get one more line. The guidelines say they should disclose "both their artificial nature and the person on whose behalf they are acting".
- An opening that does both: "Hi, you've reached the AI assistant for [your clinic]. I can take a booking request or a message for the team."
- If the caller asks whether it's a person, it says no, every time.
- Any time the caller asks, it passes them to a person or takes a message.
Who is responsible, me or the software company?
Article 50(1) puts the duty on the provider. That is the company that develops the AI system, or has it developed, and puts it into service under its own name. The guidelines give an example: a company that builds a chatbot in-house and uses it under its own name is a provider.
If you use a voice platform's agent, the platform that develops it is a provider. Whether your business counts as one too depends on how the agent was built and whose name it runs under. If that matters for you, ask a lawyer.
Either way, put the disclosure in the agent's first sentence and test it on every call path. It costs one sentence.
What should the agent never do?
- Pretend to be a person, or use a human name without saying it's an AI.
- Give medical advice, or decide how urgent a call is. Urgent callers get the emergency number.
- Confirm a booking on its own. It takes the request, and your team confirms it.
- Take card details by voice.
For what it should answer, see whether AI can answer customers' questions after hours.
How the voice mentor I run says it
- At The Brutally Honest, my own company, founders talk to an AI voice mentor in my co-founder's cloned voice.
- Since October 13, 2026, its first sentence says it's an AI voice mentor trained on his method. That opening is set in all 82 of its languages.
- Asked whether it's AI, it says yes, and that it isn't my co-founder in person.
- Every change is tested on a copy of the mentor before it reaches founders.
What it won't do
- This is my plain reading of the official texts, not legal advice. Have a lawyer check your setup.
- Outside the EU this rule may not apply. In the forum threads I read, one owner said complaints stopped once their bot said it was automated in its first sentence.
- Saying it's an AI won't save a bot that can't help. It needs a quick way to reach a person.
More questions
- Does the rule cover website chat too?
- Yes. Article 50(1) covers AI systems that interact directly with people, and the Commission names chatbots as an example.
- What if a person reviews every reply before it goes out?
- Then those replies may fall outside the rule. The guidelines leave out AI output that has been "properly reviewed and sent by humans" as the main contact. A person who could review but doesn't won't count.
- Since when does this apply?
- Since August 2, 2026, according to the European Commission's page on Article 50.
Sources
- EU AI Act, Article 50 (AI Act Service Desk, European Commission), read Oct 19, 2026
- European Commission: Transparency obligations under Article 50, FAQ (updated July 24, 2026), read Oct 19, 2026
- European Commission: Guidelines on the transparency obligations under Article 50 (July 20, 2026), read Oct 19, 2026
I build this for clients
A voice agent for your phone