Using a chatbot? What the EU AI Act says you must disclose

More and more businesses let a digital assistant answer customer questions, via WhatsApp, the chat on their website or even the phone. That's allowed, but the customer must know they're talking to AI. In this article you can read what the EU AI Act says about it, what the Dutch regulators expect, and how to get it right in practice.

What the AI Act requires

Article 50 of the EU AI Act (Regulation (EU) 2024/1689) contains transparency rules. One of the most important: AI systems that communicate directly with people must be designed so that those people know they're dealing with an AI system. That rule has applied since 2 August 2026. It does not apply if it is already obvious to a reasonably well-informed, observant person that they're talking to AI.

The regulation also says how: the information must be clear and recognisable, and the customer must receive it at the latest at the first interaction. The notice must also be accessible, so understandable for someone with a disability too.

For a trades business using a WhatsApp assistant, this means in practice that the customer is told at the start of the conversation that they're chatting with a digital assistant.

Who does it apply to?

The regulation places this duty mainly on the provider: the party that builds or supplies the AI system. But you're the one deploying the assistant, and the customer sees your business name. So choose a supplier that handles the notice by default, and check for yourself what the first message looks like. Don't assume it's "obvious anyway": an assistant that answers quickly and politely is, for many people, indistinguishable from a member of staff.

The rule applies to every AI system that talks directly to people, so not only to a chat, but also to an assistant that answers the phone.

How to get it right

• Say it in the first message, in plain language: "You're chatting with the digital assistant of [business]."
• If an AI answers the phone, say so right at the start of the call.
• Let the customer know they can also speak to a person, and how.
• Don't let the assistant make promises beyond your information.
• Keep the conversations, so you can check what was said.

What the Dutch regulators expect

On 2 October 2025, the Dutch Data Protection Authority (AP) and the Netherlands Authority for Consumers and Markets (ACM) jointly called on businesses to use chatbots in customer service with care. Their main points: people must always have the option to speak to a member of staff, it must be clear when you're talking to a chatbot, and a chatbot must not give incorrect, evasive or misleading information. An assistant that intercepts everything and never lets the customer reach a person does not fit what the regulators expect.

What if you don't arrange it?

The AI Act sets fines for not complying with the transparency rules in article 50: up to 15 million euros or 3% of worldwide annual turnover. For SMEs, the lower of those two amounts applies. For a small business, customer trust probably matters even more: someone who finds out afterwards that they were talking to a computer without knowing it doesn't feel taken seriously.

What about GDPR?

GDPR still applies alongside the AI Act. Only ask for the data you need, state in your privacy statement that you use a digital assistant, and agree with your supplier how data is processed, for example in a data processing agreement.

How Corvo Apex handles it

With the Reachable+ WhatsApp assistant the first message always says the customer is chatting with a digital assistant (AI), a member of staff can always follow along and take over, tricky questions go to you, and every conversation can be read back. This article is a general explanation, not legal advice; if in doubt, get advice.