Note. This page explains the rules as they apply to a chatbot on a website. It is not legal advice. For your own situation, ask your lawyer.
What Article 50 requires
The AI Act sorts AI systems by risk. A chatbot that answers questions about a business from that business's own content is a system with a transparency duty: allowed, as long as people know what they are dealing with. Article 50 asks two things that concern a chatbot:
- People must be told they are interacting with an AI, unless that is obvious from the context. The information has to be clear, and given at the latest at the first interaction.
- Generated text must be marked as such in a machine-readable way, so that software can tell it was produced by AI.
These duties have applied since 2 August 2026. The EU's Digital Omnibus postponed only the machine-readable marking, and only for generative systems already on the market before that date, until 2 December 2026. Telling people they are talking to an AI was not postponed.
Who is responsible
The AI Act speaks of the provider, who builds or supplies the system, and the deployer, who uses it under their own name. For a chatbot on your website, the duty to disclose sits first with the provider of the system, and you as deployer remain responsible for how you use it. In practice: choose a chatbot that discloses by itself, and do not hide or undo it.
Breaching the transparency duties can be fined under Article 99 of the Act, up to 15 million euros or 3% of worldwide annual turnover, whichever is higher, with the lower of the two for small and medium-sized businesses.
What ChatterLab builds in
- The AI notice. Every conversation opens with a line such as "You're chatting with an AI assistant", in the widget's language. No plan, add-on or setting removes it, white label included, and it sits inside the chat where the page around it cannot hide it.
- Machine-readable marking. Every answer in the widget is marked in the page's code as AI-generated.
- It never claims to be a person. Asked whether it is human, the agent says it is an AI assistant, whatever name and personality you give it.
- High-risk uses are refused. Using AI to select job applicants, to score creditworthiness or to assess people's health is high-risk under the Act and carries obligations this product does not meet, so an agent declared for recruitment, credit scoring or healthcare is not created.
- Records. Every conversation is kept with the sources behind each answer, and an audit log records what changed and who changed it.
If you build your own chat on the API, none of this is drawn for you, and telling your users they are talking to an AI becomes your job.
A checklist for your website chatbot
- The chat says it is an AI before or at the first message, in a language your visitors understand.
- Nothing on your page covers or removes that notice.
- The chatbot does not pretend to be a person, by name or by tone.
- It is not used to assess people: their health, their creditworthiness or their suitability for a job.
- Your privacy notice names the chat and what it stores. The GDPR page lists what you need.
- You can show afterwards what it said, and why.
The EU AI Act in our documentation goes into each point.
Questions about the AI Act
Does the AI Act apply to my website chatbot?
Yes, if it talks to people in the EU. A customer-service chatbot is not high-risk, but it falls under the transparency duty of Article 50: people must be told they are talking to an AI.
Since when do chatbots have to say they are an AI?
Since 2 August 2026. The Digital Omnibus postponed only the machine-readable marking of generated content, for systems already on the market before that date, until 2 December 2026.
Can I remove the AI notice from the ChatterLab widget?
No. It stays on every plan and with every add-on, white label included, because the law asks for it.
Is a chatbot for a clinic or a recruiter allowed?
A clinic's chatbot that gives opening hours and explains how to book is customer service. One that assesses symptoms, creditworthiness or job applicants is high-risk, and ChatterLab refuses to create an agent declared for those uses.
What are the fines?
Breaching the transparency duties of Article 50 can be fined up to 15 million euros or 3% of worldwide annual turnover, whichever is higher. For small and medium-sized businesses the lower of the two applies.