Your Chatbot Has to Say It's a Chatbot

There's a chat widget on your site, or a tool writing your product copy, and someone has asked whether that's a problem in Europe now.

I went through Article 50 and the Commission's own Q&A on it, line by line, then ran the checks on my own site.

The intake chat on daisyguti.ai is one I had built and run under my own name, so the provider duties are mine on that one. My About page calls it "the AI assistant that takes your intake." Inside the widget, every message carries the label "Daisy's assistant."

Article 50 — the transparency part of the EU AI Act, about telling people a machine is involved — applies from 2 August 2026 (European Commission). Most of what it asks for is owed by whoever built the tool. You owe a narrow slice of it.

Here's how to tell which you are.

Does this reach you at all

Article 50's reach is set by Article 2. Three triggers matter:

• Article 2(1)(a) — providers placing an AI system on the market or putting it into service in the Union, established there or not.
• Article 2(1)(b) — deployers established or located in the Union.
• Article 2(1)(c) — providers and deployers in a third country, where the output produced by the AI system is used in the Union.

Read (c) twice.

Having EU customers is not the trigger. Selling to someone in Berlin does not pull your US business in. Placing an AI system on the EU market or into service there does, and so does your system's output being used in the Union.

Your analytics answer that side of it. In Search Console for daisyguti.ai, 9 May to 6 August 2026, five EU countries show impressions — Germany, the Netherlands, Spain, Hungary, Lithuania — ten in total, no clicks. That covers Google search only, so Instagram and direct traffic sit outside it.

Provider or deployer

The Commission's Q&A on Article 50 draws the line:

Provider:

…persons, public authorities, agencies or other bodies that develop AI systems, or have them developed, and place them on the EU market or put them into service under their own name or trademark.

Deployer:

…using AI systems under their authority, excluding use for personal, non-professional activities.

If someone else built the tool and you're the one running it, you're a deployer. Had one built and put your own name on it — read the provider duties as yours.

What your vendor owes

Two duties, both on providers.

Article 50(1) — build it so people know. The provider designs the system so anyone interacting with it directly is informed they're dealing with an AI system. It's a design duty, so the disclosure ships inside the product.

Article 50(2) — mark the output. Providers of systems generating synthetic audio, image, video or text ensure those outputs are marked in a machine-readable format and detectable as artificially generated or manipulated, by technical means that work as far as is technically feasible.

Nothing in 50(2) tells you to watermark your own posts. Whoever supplies the generating system carries that one.

There's also a Code of Practice on Transparency of AI-generated Content, published 10 June 2026. Adherence is voluntary, and the Commission notes the Article 50 requirements are legal obligations either way.

What you owe

Deepfakes — Article 50(4), first subparagraph. A deployer of a system that generates or manipulates image, audio or video content constituting a deep fake discloses that the content is artificially generated or manipulated.

The Commission's Q&A is specific about how: the provider's embedded technical mark doesn't carry it. The disclosure has to be perceivable by a person, "e.g. with visible or audible labels."

Text published to inform the public on matters of public interest — Article 50(4), second subparagraph. A deployer publishing AI-generated or manipulated text for that purpose discloses it.

That's narrower than "anything you publish." A product description isn't it. A post about a local election or a public health question is where the duty bites.

Article 50(3) adds one more, outside most small-business setups: run an emotion recognition or biometric categorisation system and you inform the people exposed to it.

Article 50(5) sets the timing for 50(1) through 50(4): the information reaches the person in a clear and distinguishable manner at the latest at the time of first interaction or exposure, and conforms to applicable accessibility requirements.

The exceptions, each with the duty it belongs to

The carve-outs don't travel between duties.

Duty	Its exception	What that doesn't do
50(1) tell people it's an AI system	Unless it's obvious from the point of view of a natural person who is reasonably well-informed, observant and circumspect, given the circumstances and context of use	Doesn't reach 50(2) or 50(4)
50(2) machine-readable marking	Where the system performs an assistive function for standard editing, or doesn't substantially alter the deployer's input data or its semantics	Doesn't excuse a deployer's 50(4) disclosure
50(4) deepfake disclosure	Where the content is part of an evidently artistic, creative, satirical or fictional work, the obligation is limited to disclosing that the generated content exists, in a manner that doesn't hamper display or enjoyment of the work	Doesn't remove the disclosure. It shapes how it's made
50(4) public-interest text	Where the content has undergone human review or editorial control and a natural or legal person holds editorial responsibility for the publication	Both halves are required

Each paragraph also carves out use authorised by law to detect, prevent, investigate or prosecute criminal offences. Article 50(6) adds that none of this displaces other transparency obligations in Union or national law.

What to check this week

1. Open your site in a private window and start a chat. Does the first message say it's an AI assistant, before you type anything? Your vendor owes that under 50(1) — check it didn't get switched off in your settings.
2. List every AI tool touching a customer. Chat widget, review reply generator, product description writer, on-site search, the voice on your phone line.
3. Mark each one provider or deployer.
4. Find anything you publish that could pass for a real recording of a real person. That's 50(4). Label it where a person sees it, at first exposure.
5. Separate public-interest writing from marketing. If a tool drafted something published to inform the public, disclose it — or have a named person review it and hold editorial responsibility.
6. Ask your vendor, in writing: how do you mark generated output as machine-readable and detectable under Article 50(2)? Keep the answer.

What this guide doesn't cover

Article 50 only. The AI Act also has prohibited practices (Article 5), high-risk classification (Article 6 and Annex III), obligations for general-purpose AI models, and a conformity assessment regime with CE marking and notified bodies. None of that is in here, and neither is GDPR, the Data Act, or any US state law.

The Commission's guidelines on this one article, adopted 20 July 2026, are advisory. They cover edge cases this guide skips.

This is general information. I'm an engineer, and legal advice is an attorney's job. For whether a specific system of yours is caught, and what to write on the page, ask a licensed attorney.

Start Here

At daisyguti.ai/work-with-me there's an AI intake assessment that maps where your business stands before any call. It takes a few minutes and gives you a clear read on what to hand off first.

Sources

1. European Commission, "Guidelines on transparency obligations for AI systems" — https://digital-strategy.ec.europa.eu/en/policies/guidelines-ai-transparency-obligations
2. EU AI Act, Article 50 (Transparency obligations for providers and deployers of certain AI systems) — https://artificialintelligenceact.eu/article/50/
3. EU AI Act, Article 2 (Scope) — https://artificialintelligenceact.eu/article/2/
4. European Commission, Q&A on transparency obligations under Article 50 — https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act
5. European Commission, "Guidelines on transparency obligations for providers and deployers of AI systems," adopted 20 July 2026 — https://digital-strategy.ec.europa.eu/en/library/guidelines-transparency-obligations-providers-and-deployers-ai-systems
6. European Commission, Code of Practice on Transparency of AI-generated Content, published 10 June 2026 — https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content