EU AI Act chatbot rules: what Article 50 requires and when

Article 50(1) says providers must make sure that AI systems intended to interact directly with natural persons are designed and developed so that the people concerned are informed that they are interacting with an AI system. The exception is when this is obvious to a reasonably well-informed, observant and circumspect person, given the circumstances and the context of use. The duty doesn't apply to AI systems authorised by law to detect, prevent, investigate or prosecute criminal offences, subject to appropriate safeguards.

Article 50(5) covers how and when people must be told. The information must be given in a clear and distinguishable manner at the latest at the time of the first interaction or exposure, and it must meet the applicable accessibility requirements.

Article 50 has other transparency duties too. Only the first one in this list is specific to chatbots:

• Article 50(1): providers of systems that interact directly with people must design them so people know they're dealing with AI.
• Article 50(2): providers of systems that generate synthetic audio, image, video or text must mark the outputs in a machine-readable format so they can be detected as AI-generated.
• Article 50(3): deployers of emotion recognition or biometric categorisation systems must inform the people exposed to them.
• Article 50(4): deployers must disclose deepfakes and AI-generated text published to inform the public on matters of public interest. For that text, the duty doesn't apply when it has gone through human review and someone holds editorial responsibility.

Article 3(3) defines a provider as a person or body that develops an AI system, or has one developed, and places it on the market or puts it into service under its own name or trademark, whether for payment or free of charge. Article 3(4) defines a deployer as anyone using an AI system under its authority, except for personal, non-professional use.

The chatbot disclosure in Article 50(1) is a provider duty. The labelling duties in Article 50(3) and 50(4) are deployer duties. If your company builds a support assistant and runs it under its own brand, read both definitions carefully against how the system was built and who puts it in front of customers.

The Commission's FAQ on Article 50 says its guidelines set four cumulative criteria for the interaction duty: the system is an AI system, it is built for a genuine two-way exchange, it communicates directly rather than through a human intermediary, and it interacts with natural persons. The FAQ says systems that work only in the background, through machine-to-machine communication or without direct contact with people, fall outside this duty. It also says the "obvious" exception should be read restrictively.

In practice, many teams start each AI conversation with a short line such as "You're chatting with an AI assistant. Type 'agent' to reach a person." Doing this is a design choice. Whether it's enough for your system is a question for the guidelines and your adviser.

Article 113 sets the AI Act's timeline. In November 2025, the Commission proposed changing parts of it in a Digital Omnibus on AI (COM(2025) 836, dated 19 November 2025). That proposal has since been adopted: after a political agreement on 7 May 2026 and formal adoption, it was published as Regulation (EU) 2026/1744. The Commission says it entered into force on 27 July 2026. The table shows the adopted dates.

One difference between the proposal and the adopted text affects generative AI. The Commission's proposal would have given providers of generative systems already on the market a six-month grace period for Article 50(2) marking. The adopted regulation sets the deadline at 2 December 2026. According to the Commission's FAQ, the chatbot disclosure in Article 50(1) still applies from 2 August 2026.

Under Article 99(4), breaching the Article 50 obligations can lead to administrative fines of up to EUR 15,000,000 or, for an undertaking, up to 3% of its total worldwide annual turnover for the preceding financial year, whichever is higher. For SMEs, including start-ups, Article 99 applies whichever of the two amounts is lower.

On 20 July 2026, the Commission published guidelines on the Article 50 transparency obligations. A separate, voluntary Code of Practice on Transparency of AI-generated Content covers Article 50(2), (4) and (5). The Commission says not signing it doesn't mean non-compliance. For the chatbot duty in Article 50(1), the FAQ says providers and deployers can decide on suitable measures themselves, taking the guidelines into account.

• Put the AI disclosure in the first message or the voice greeting, not only in your terms of service.
• Make the disclosure readable by screen readers and clear in voice, since Article 50(5) mentions accessibility.
• Give customers a visible way to reach a person, and pass the conversation history along so they don't have to repeat themselves.
• Keep a record of the disclosure text and when you changed it, so you can show what customers saw.

TascaLab builds chat and voice bots on top of the tools you already use (Zendesk, Freshdesk, Jira, Slack and internal tools). Standard Chat Bot projects cost $575 and Voice Bot projects $1,350. Both include basic documentation and 30 days of post-launch support.