← Back to the blog
AI agents

Your chatbot must say it is AI. What the EU AI Act requires since August

Since 2 August, the EU AI Act requires chatbots to make clear that customers are talking to a machine. It also reaches companies outside the EU whose chatbots serve people in Europe.

Speech bubble with a highlighted notice in dark lines on a light background, beside the glass form with a blue core

If your company serves customers with a chatbot on its website or on WhatsApp, and any of them are in Europe, a new rule already applies to you.

What the rule says

Article 50 of the EU AI Act has applied since 2 August 2026. It asks for three main things:

  • Chatbots must disclose that they are AI, unless that is obvious to a reasonably attentive person.
  • Audio, images and video that imitate real people, such as a cloned voice, must be labelled as generated or altered by AI.
  • AI-generated text published to inform the public on matters of public interest must also be labelled, unless it has gone through human review.

Fines can reach 15 million euros or 3% of worldwide annual turnover, whichever is higher, according to Cooley's analysis.

Does it apply outside the EU?

Yes, when the chatbot serves people in the European Union. The law covers systems whose output is used in the EU, wherever the company is based. That includes UK and US businesses selling into Europe.

The "obvious" exception is narrower than it sounds. According to Travers Smith, the European Commission takes the view that a support chatbot that could be mistaken for a person must disclose.

Support flow in which the customer opens the chat, is told they are talking to an AI, the agent answers and hands over to a person when the customer asks

Who is responsible

The law separates whoever builds the system from whoever uses it. The vendor of the chatbot tool must make disclosure possible. The company that puts the chatbot in front of its own customers must make sure it actually appears.

In practice, even with an off-the-shelf tool, check that the notice shows up and that your vendor contract covers it.

What to change

  1. A clear first message, such as "You are chatting with our virtual assistant", on the website and on WhatsApp.
  2. No name and photo pretending to be a person. The agent can have a name, as long as it does not hide that it is AI.
  3. An easy way to reach someone on the team, whenever the customer asks.
  4. A spoken notice too, if support runs by phone or voice.
  5. Labels on content with cloned voices or AI-generated people, such as product videos and dubbed courses.

Beyond compliance, saying it is AI builds trust. Customers feel better when they know who they are talking to.

How we work

In the WhatsApp AI agents we build, the notice that it is AI comes in the first message, and the customer can ask for a person at any time. For courses dubbed in the author's own voice, we agree with you how to tell learners that AI was used.

This article is for information only and is not legal advice for your situation.