Since 2 August 2026, Article 50(1) of the EU AI Act has applied: AI systems that interact directly with people must be designed so that those people are informed they are interacting with an AI system. It is the most visible AI Act duty, because your customers see it, and one of the easiest to get right.
Who carries the duty
The text puts it on the provider: the company that builds the system and puts it out under its own name. In practice there are two common situations:
- You built the assistant yourself on a model API and run it on your site or in your product. You are the provider. The notice is yours.
- You use a chatbot platform and configured it for your website. The platform is the provider, and most platforms give you a disclosure setting. The legal duty sits with them, but you control what visitors see. If you switch the notice off, rename the bot “Laura from Customer Care” and let it pass as a person, you have created the exact situation the rule exists to prevent, and consumer-protection law in most EU countries does not look kindly on it either.
Either way the practical answer is the same: make sure the notice is there. If you are not sure which side you are on, our note on provider or deployer walks through it.
When it is “obvious” and you can skip it
The Act exempts cases where it is obvious from the point of view of a reasonably well-informed, observant and circumspect person, taking into account the circumstances and the context of use. A tool clearly sold as “AI writing assistant” probably qualifies. A support chat with a human name and a stock photo does not. When in doubt, add the notice: it costs one sentence.
How and when to say it
Article 50(5) asks for the information to be given in a clear and distinguishable manner, at the latest at the first interaction, and in a way that meets accessibility requirements. What that looks like by channel:
| Channel | What works |
|---|---|
| Website chat | First message: “Hi, I’m the [Company] AI assistant.” A visible “AI” label next to the name, not only in the terms |
| Voice assistant or phone bot | Said at the start of the call, before anything else: “You are speaking with an AI assistant.” |
| Email or messaging agent | A line in the first reply and in the signature: “This reply was written by our AI assistant.” |
| In-product assistant | An “AI” badge on the panel and in the first response, not in a tooltip |
Three habits make it stick:
- Offer a way to a human where the conversation matters (complaints, cancellations, health, money). Not required by Article 50, but it is what people expect once they know.
- Keep the notice when you redesign. Disclosures disappear in redesigns more often than through any decision.
- Write down what you show and where. One screenshot per channel in a shared folder is evidence enough to answer a client or an authority.
Related duties that sit next to it
Article 50 has other parts that often apply to the same company: marking AI-generated content (providers, with extra time until 2 December 2026 for systems already on the market), labelling deepfakes, and disclosing AI-generated text published on matters of public interest. A chatbot notice is usually the first item on the list, not the only one.
Sources. Regulation (EU) 2024/1689, Article 50(1) and 50(5), as amended by Regulation (EU) 2026/1744 (Digital Omnibus on AI). Dates verified on 5 October 2026.
A practical readiness and implementation engagement. It does not replace your legal counsel, a formal conformity assessment or certification. We work alongside your counsel, never in its place.