Do I have to disclose that an AI is answering? What EU law requires
MHBy Manuel Hdez
Do you have to disclose that an AI is answering? Yes. Since August 2, 2026, the EU AI Act requires that anyone talking to an AI assistant knows it's an AI, unless that's obvious. Saying it clearly in the first message is enough. You don't need to sound like a robot, you just can't pass as a person.
What exactly does the law say?
It's Article 50 of Regulation (EU) 2024/1689, the AI Act. AI systems meant to talk directly with people have to be built so that the person knows they're talking to an AI, unless it's obvious to a reasonably attentive person. The notice has to be clear and arrive, at the latest, at the first interaction.
And it wasn't postponed. The simplification package approved this summer (the so-called Digital Omnibus) delayed the high-risk AI obligations, but left this one on its original date. The only extension inside this article was for the technical marking of AI-generated content, which is a different matter.
Whose job is it, yours or whoever builds it?
The rule puts the obligation on the provider, the one who designs the system. But note one detail: the law also treats as a provider anyone who commissions a system and puts it into service under their own name or brand. An assistant another company builds for you and that answers in your business's name can leave you as the responsible party.
In practice it doesn't matter who signs: the one who looks bad in front of the customer is you. If someone builds it for you, ask for the notice to come built in. And if your case has any special twist, check with a lawyer; this is an explanation, not legal advice.
How do you say it without putting customers off?
With one sentence at the start, in your business's tone, and a clear way to reach a person. For example: 'Hi, I'm the school's automated assistant. I can tell you schedules, prices and free places. If you'd rather talk to Marta, type PERSON and she'll answer you'.
That doesn't scare anyone off. What does is finding out halfway through a conversation that there was no person on the other side. Saying it upfront does the opposite: the customer knows what to expect and how to reach you.
What if it's already obvious it's a bot?
The 'if it's obvious' exception exists, but it's not worth leaning on. A well-built assistant looks a lot like a person, which is exactly what you want. The sensible thing is to always disclose: it costs one line and removes the doubt.
The law doesn't change the rest either. It doesn't ban automation or force you to sound mechanical, and data protection still applies the same way to whatever the assistant collects.
Frequently asked questions
Is it mandatory to disclose that a chatbot is an AI?
Yes. Since August 2, 2026, Article 50 of the EU AI Act requires that people know they're talking to an AI system, unless it's obvious to a reasonably attentive person.
Did the Digital Omnibus postpone this obligation?
No. The Digital Omnibus delayed the high-risk AI obligations, but the duty to disclose that people are talking to an AI has applied since August 2, 2026, as planned. Only the technical marking of AI-generated content got an extension.
How do I have to tell my customers?
Clearly and, at the latest, at the first interaction. One sentence at the start of the conversation saying it's an automated assistant, what it can do and how to reach a person is enough for most businesses.
Does it also apply to WhatsApp and voice assistants?
Yes. It applies to any AI system that talks directly with people: the chat on your website, a WhatsApp assistant or one that answers the phone. What counts is that the person interacts with the AI, not the channel.
Sources
If the assistant will live on WhatsApp, first find out whether your business needs the WhatsApp Business app or the API.
Want to see this applied to your business? Let's talk.
Free diagnosisKeep reading
