Since 2 August 2026, Article 50 of the EU AI Act applies. It affects many business websites: visitors who chat with an AI system must be told it is an AI. Below: what the official sources say, and a one-minute check for your own site.
Short answer: if your chatbot uses AI and holds real conversations, it should say it is an AI, clearly and at the start of the first conversation, unless that is obvious to an average visitor. A line in your terms and conditions is not enough.
What Article 50(1) requires
Article 50(1) requires providers to design AI systems that interact directly with people so that those people "are informed that they are interacting with an AI system". Article 50(5) adds how: "in a clear and distinguishable manner at the latest at the time of the first interaction", meeting applicable accessibility requirements.
The Commission's FAQ on Article 50 (last updated 24 July 2026) and its guidelines on Article 50 list four conditions:
- it is an AI system (rule-based quick replies and classic out-of-office emails are not);
- it is built for a genuine two-way exchange, not just collecting data;
- the AI itself talks to the person, not through a human employee;
- the other side is a natural person.
The guidelines name customer support and e-commerce chatbots and AI voice assistants as examples in scope.
The "obvious" exception is narrow
No notice is needed if it is obvious to a reasonably well-informed, observant and circumspect person that they are talking to an AI. The FAQ says this exception "should be interpreted in a restrictive manner". The guidelines limit it to cases with "almost no doubt left", and list helpdesk chatbots whose replies users "may perceive as human-generated" as needing the notice.
Who has to do it: provider or deployer?
Article 50(1) is addressed to the provider: whoever develops an AI system, or has it developed, and puts it on the market or into service under its own name or trademark. A deployer uses an AI system under its authority.
- You built a chatbot, or had one built, and run it under your brand: you are likely the provider (the guidelines' example: an in-house chatbot under the company's own name). Modifying an existing AI system, for example with new training data, and running it under your name also counts.
- You use a vendor's chatbot as-is: the vendor is the provider, you are a deployer. But the notice appears in your chat window, often via settings you control, so agree who sets it.
Since when, and did the Digital Omnibus change it?
Article 50 applies from 2 August 2026 (Article 113). The Digital Omnibus on AI, Regulation (EU) 2026/1744, published on 24 July 2026, did not move that date for chatbots. It only gives generative systems already on the market until 2 December 2026 for machine-readable marking of AI-generated content (Article 50(2)). The guidelines confirm the chatbot notice "must be ensured as of 2 August 2026", also for systems that went live earlier.
Enforcement is mainly up to national market surveillance authorities. Article 99 sets fines for breaching Article 50 of up to €15 million or 3% of worldwide annual turnover, whichever is higher. For SMEs and start-ups the cap is whichever is lower.
The 1-minute check
Use your phone first, then desktop.
- Be a new visitor. Private window, open the chat, type nothing yet.
- Is the notice there before your first question? Inside the chat window: a first-turn greeting or a label such as "You are interacting with an AI system", ideally near the input field.
- Is it plain and specific? "I'm an AI assistant, not a human" works. Not enough on their own, per the guidelines: a vague "assistant", a site-wide "Services on this website use AI", or "this system uses LLMs".
- Is it hard to miss? A notice only in your terms, privacy policy or a menu is not clear. Nor is tiny grey text hidden by the phone keyboard.
- Is it real text? Good contrast, readable by a screen reader. An "AI" badge can support the text, not replace it.
- Do the name and photo fit? A "Lisa" with a real person's photo makes the AI less obvious.
- Ask it. Type "Am I talking to a human?" The guidelines expect the system to disclose its AI nature when asked.
Also check
- Voice and phone bots: a spoken line at the start, such as "This is an AI-powered assistant". A tone alone is not enough.
- AI that sends emails: the guidelines' example is an AI label at the top of the email.
- Human in the loop: AI drafts that a person reviews and sends as the main contact are outside Article 50(1). AI replies that go out directly need the notice.
What a passing notice looks like
The AI Act prescribes no fixed wording. A typical fix is a welcome message such as "Hi, I'm the AI assistant of [company]. I'm not a human, but I'm happy to help." plus a short "AI assistant" label next to the input field. In many chat tools you can edit this yourself.
This is technical help, not legal advice. If your bot handles health, financial advice or complaints, or talks with children, the guidelines say periodic reminders are likely to be necessary.
Free check or fixed-price fix
Request a free check at foreground.info/nl/ai-check (the page is in Dutch). You get a pass/fail with a screenshot within 24 hours.
Or have it done: An AI notice for your chatbot, €390 fixed, 1 day. I check your chatbot and add a clear notice, with screenshots before and after.
Work with me
I'm Marc de Ruijter, a Dutch full-stack developer with 25+ years in IT. I work remotely for EU clients at fixed prices, in Dutch or English: websites, web apps and AI features such as chat assistants. See my offers on the hire page or send me a message.