AI Chatbot Disclosure Requirements EU: What Article 50(1) Actually Says
By AJ Fox, A50 Check · Vendor and regulation details verified August 26, 2026
If your website has a chat widget that answers visitors automatically, EU law now requires you to tell people they're talking to a machine — and a cookie banner doesn't count. Article 50(1) of the EU AI Act took effect on August 2, 2026, and it applies whether your business is based in Berlin or Bloomington, Indiana, as long as EU visitors can reach your chatbot.
This article breaks down exactly what the rule requires, who it covers, the most common ways businesses get it wrong, and how to fix a missing disclosure in about ten minutes.
What Are the AI Chatbot Disclosure Requirements EU Businesses Must Follow?
The core rule sits in Article 50(1) of Regulation (EU) 2024/1689 (the EU AI Act). The full text of Article 50 says:
"Providers shall ensure that AI systems intended to interact directly with natural persons are designed and developed in such a way that the natural persons concerned are informed that they are interacting with an AI system, unless this is obvious from the point of view of a natural person who is reasonably well-informed, observant and circumspect, taking into account the circumstances and the context of use."
In plain English: if a visitor is chatting with something and it's not obvious they're talking to software, you have to tell them. There's no revenue threshold, no company-size exemption, and no carve-out for small businesses in Article 50(1) itself — it applies to essentially anyone deploying an AI chatbot, voice agent, or virtual assistant that talks directly to people (see this practical guide to the transparency rules).
Article 50(5) adds a timing rule: the disclosure has to be "clear and distinguishable" and provided "at the latest at the time of the first interaction," and it has to meet accessibility standards — meaning it can't be buried in tiny gray text or hidden behind a click (Article 50 text; analysis of the "obvious" exemption).
Who Does This Apply To?
Two roles matter under the AI Act: providers (the company that builds the AI system, like Intercom or Zendesk) and deployers (the business that puts it on their website — probably you). Article 50(1) places the design obligation on providers, but in practice, most chatbot vendors ship a default label or introductory message, and it's the deployer's job to make sure that disclosure is actually turned on and visible. Intercom's AI Agent disclosure documentation is a good example of how vendors hand that control to you.
Location doesn't get you out of it either. The obligation applies to non-EU companies if their chatbot's output reaches people in the EU — so a US-based e-commerce site or SaaS product with European customers is in scope even if the business itself has no EU office (per the European Commission's transparency guidelines).
What Doesn't Count as Valid AI Chatbot Disclosure
This is where most sites fail without realizing it. A few things people assume satisfy the rule but don't:
- A cookie consent banner. Cookie banners are about data privacy (GDPR), not AI interaction. They say nothing about whether you're talking to a bot.
- A generic "Terms of Service" mention buried on a separate page. The law requires disclosure "at the latest at the time of the first interaction" — a document nobody reads during checkout doesn't meet that bar.
- A tiny "Powered by [Vendor]" logo in the corner. This tells a user which software runs the chat, not that they're talking to AI rather than a human.
- Assuming it's "obvious." The obvious-from-context exemption is narrow and judged from the perspective of a "reasonably well-informed, observant and circumspect" person — regulators are not likely to accept "everyone knows chatbots are AI now" as a blanket excuse, especially for chatbots styled to look human (photo avatars, first-person "I" language, human names with no AI label).
Common Misconceptions About AI Chatbot Disclosure Requirements EU-Wide
Misconception 1: "My chatbot is simple rules-based logic, not real AI, so it doesn't count." If the system uses machine learning, natural language processing, or generative AI to interact with users — even a "simple" GPT-powered widget — it likely qualifies as an AI system under the Act's broad definition. Basic decision-tree bots with no learning component are a grayer area, but most modern chat widgets (Intercom Fin, Tidio Lyro, Zendesk AI agents) are squarely in scope.
Misconception 2: "I already have a privacy policy that mentions AI." A privacy policy explains data handling; it doesn't satisfy the "clear and distinguishable" real-time disclosure requirement at first interaction.
Misconception 3: "This is only about deepfakes and fake news." That's Article 50(4), a separate and narrower rule about synthetic media and public-interest text. Article 50(1) — the one about chatbots — is much broader and applies to nearly every business with an AI-powered chat or voice interface. We cover the difference between these two rules in detail in our companion piece on EU AI Act chatbot registration misconceptions.
Misconception 4: "There's a grace period, so I have time." Article 50 has applied since August 2, 2026, per the Commission's guidelines. There is a separate, narrower grace period (until December 2, 2026) for providers' machine-readable marking of pre-existing generative AI systems — but deployer disclosure duties for chatbots started immediately.
What Happens If You Ignore AI Chatbot Disclosure Requirements
Non-compliance with Article 50's transparency obligations can bring administrative fines of up to €15 million or 3% of a company's total worldwide annual turnover for the preceding financial year, whichever is higher (European Commission quick facts; penalty overview). For context, that's a materially higher ceiling than most people expect for what looks like a "just add a label" fix — the size of the number reflects that the EU treats transparency as a baseline right, not a nice-to-have.
In practice, enforcement in the early period is more likely to start with a complaint or audit than a surprise raid, and regulators generally look at intent and effort. A business that has already added an AI label but worded it awkwardly is in a very different position than one that never addressed it at all. Still, the exposure is real, and it scales with company revenue — not with how big your chatbot budget was.
How to Fix AI Chatbot Disclosure Requirements in About 10 Minutes
You don't need a legal team to close the most common gap. Here's a fast checklist:
| Step | What to do | Time |
|---|---|---|
| 1 | Open your chatbot vendor's settings and find the "AI label" or "AI Agent" toggle | 1 min |
| 2 | Turn it on if it's off — check current default status, since some vendors (like Intercom Fin) changed their default after July 2025 | 1 min |
| 3 | Edit the bot's opening message to state plainly: "You're chatting with an AI assistant" | 2 min |
| 4 | Make sure the label is visible on-screen, not just in a settings panel or footer link | 2 min |
| 5 | Check any voice agent or AI phone line for an equivalent verbal disclosure at call start | 2 min |
| 6 | Screenshot the live disclosure for your own compliance records | 2 min |
If your chatbot vendor doesn't offer a built-in label, you can usually add one manually by editing the bot's greeting script or the widget's header text — most platforms allow this without developer help.
Check your own site: run a free scan at a50comply.com — it takes about 60 seconds, no email required, and it flags the AI chat widgets and voice agents your pages are actually loading.
FAQ
Does every website with a chatbot need an AI disclosure under EU law? Almost every site where a chatbot is not obviously a bot needs some form of disclosure under Article 50(1), regardless of company size, as long as EU users can interact with it.
Is a cookie banner enough to meet AI chatbot disclosure requirements? No. Cookie banners address data privacy under GDPR, not AI interaction transparency, and they don't satisfy Article 50(1) at all.
Does this apply to US companies with no EU office? Yes, if your chatbot's output reaches people located in the EU, the obligation applies regardless of where your company is headquartered.
What counts as "clearly disclosing" AI use? A plain-language statement, visible at or before the first interaction, that a user is talking to an AI system — not a hidden policy page, not a vendor logo, and not an assumption that it's "obvious."
Do voice agents and phone bots need disclosure too? Yes. Article 50(1) covers any AI system "intended to interact directly with natural persons," which includes voice assistants and automated phone systems, not just text chat.
Related reading
- Which Website Tools Secretly Use AI in 2026 (And Need EU Disclosure)
- EU AI Act Chatbot Registration: Why Almost No One Actually Needs It
- How the A50 Check scanner works
Run a free scan at a50comply.com to see if your site's AI chatbots, chat widgets, and voice agents are exposed.
This article is general information, not legal advice. It does not create an attorney-client relationship and should not be relied on as a substitute for advice from a qualified lawyer familiar with your specific business and jurisdiction.