EU AI Act & Chatbots: What the New Transparency Rules Mean for Your Business

If your website runs an AI chatbot, a new EU-wide transparency obligation applies from 2 August 2026. Article 50 of the EU AI Act requires that people know when they're talking to a machine rather than a person. This guide explains what the rule actually says, which deadlines matter, and what a small or medium-sized business realistically needs to do about it. It is general orientation, not legal advice — for your specific situation, speak to a qualified lawyer.

What Article 50 actually requires

Article 50(1) of the EU AI Act requires providers of AI systems that interact directly with people — a chatbot on your website is the textbook example — to design the system so that the people interacting with it are informed they are dealing with an AI, unless this is already obvious from the context. For a chatbot, that disclosure has to happen at or before the first interaction: in practice, before or right at the start of the conversation, not buried three clicks deep in a privacy policy.

This isn't a vague best-practice suggestion. It's a binding obligation under EU law, and it applies regardless of which country in the EU your visitors are browsing from, because the AI Act is a regulation, not a directive that each member state implements differently.

The key date: 2 August 2026

Article 50 becomes applicable on 2 August 2026. There has been discussion in Brussels about a broader 'Digital Omnibus' package that would simplify or delay parts of the AI Act, but it does not push back Article 50 — the transparency obligations for chatbots and AI-generated content stay on the original timeline. If you have a chatbot live on your site, or you're planning to launch one, 2 August 2026 is the date to plan around, not a date to wait and see about.

Do the rules apply to generated content too?

Yes — but with a separate, slightly more forgiving timeline. Article 50(2) requires providers of generative AI systems to mark synthetic output (text, audio, images, video) as AI-generated in a machine-readable format, so it can be detected as artificial rather than human-made. There's a transitional carve-out: generative systems that were already on the market before 2 August 2026 get until 2 December 2026 to implement this labelling. Systems placed on the market on or after 2 August 2026 don't get that grace period — they need to mark synthetic output from day one.

Other Article 50 obligations at a glance

• Article 50(3): deployers of emotion recognition or biometric categorisation systems must inform the people exposed to them.
• Article 50(4): deep fakes and AI-generated text published on matters of public interest must be disclosed as artificially generated or manipulated.
• The European Commission has draft guidelines on Article 50's scope in progress, and a code of practice on AI-generated content is being developed — both are worth watching if your chatbot does more than answer routine questions.

What happens if you don't comply

Non-compliance with the AI Act's transparency obligations can trigger fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher. There is a specific rule for smaller organisations, though: under Article 99(6), SMEs and start-ups face whichever of the two amounts is lower, not higher — the opposite way round from the default rule for larger companies. The AI Office and national market surveillance authorities gain the formal power to enforce these rules from 2 August 2026, the same date Article 50 itself starts applying. Enforcement is unlikely to focus first on a small business with a well-labelled chatbot, but 'unlikely to be first' is not the same as 'exempt' — the obligation applies to any provider, regardless of size.

A practical checklist for website chatbots

• Add a clear, visible disclosure that visitors are talking to an AI assistant — ideally in the first message of the conversation, not just in a footer link.
• Don't assume a 'chat bubble' icon is disclosure enough on its own — Article 50 asks that people are actually informed, in a way that is clear before or at the point of interaction.
• If your chatbot generates written responses dynamically (most AI chatbots do), check whether machine-readable labelling of that output applies to your setup and by which deadline.
• Review your chatbot's disclosure wording well before 2 August 2026 — don't leave it to the last week.
• Keep an eye on the European Commission's forthcoming guidelines on Article 50's scope, since some of the finer detail is still being clarified.
• If your chatbot is used for anything beyond routine business queries — for example anything touching emotion recognition or biometric categorisation — get specific legal advice, because Article 50(3) brings extra obligations.

None of this is legal advice, and the Commission's guidelines on Article 50's exact scope are still being finalised. If your chatbot handles anything sensitive — health-adjacent questions, biometric data, or content meant for public distribution — treat this article as a starting point for a conversation with a qualified lawyer, not as a substitute for one.