AI agent disclosure (EU AI Act, Article 50)
Compliance and transparencyArticle 50 transparency obligation · AI interaction disclosure · chatbot disclosure duty
The EU obligation that a person knows, from the very first interaction, that they're talking to an AI system rather than a human. It applies expressly to AI agents and avatars. The «unless it's obvious» carve-out exists — and is to be read narrowly.
Article 50 of the AI Act (Regulation (EU) 2024/1689) applies from 2 August 2026. Paragraph 1 requires providers to design systems that interact directly with people — the Commission explicitly names chatbots, AI agents and avatars — so that the person knows they are dealing with an AI. Four cumulative criteria trigger it: the system qualifies as an AI system; it is designed for a genuine two-way exchange rather than merely collecting data or emitting automated responses; the interaction is direct, with no human intermediary; and it is with natural persons, consumers or professionals alike. Systems operating purely in the background or machine-to-machine fall outside. Notice must be given from the start of the first interaction, clearly and distinguishably, and in line with accessibility requirements. The obviousness exception is measured against an average person who is reasonably well-informed, circumspect and observant, and the Commission asks for it to be read restrictively because it deprives people of transparency. Penalties reach EUR 15 million or 3% of total worldwide annual turnover for the preceding financial year. Mind the grace period: it covers only the Article 50(2) marking duty, only for systems placed on the market before 2 August 2026, and it runs out on 2 December 2026. The 50(1) disclosure gets no extension.
How it differs from
- Synthetic content marking (Art. 50(2))
- 50(2) marks the output in machine-readable form; 50(1) tells the person their counterpart is an AI.
- Deepfake labelling (Art. 50(4))
- 50(4) falls on the deployer and covers published content; 50(1) falls on the provider and covers the interaction.
- High-risk requirements (Annex III)
- Article 50 applies whether or not the system is high-risk.
Examples
- Identifying the agent as AI in the opening message of a support chat
- Keeping the notice when the agent serves professionals, not just consumers
- Documenting why a case was judged «obvious», in case you have to defend it
FAQ
- Can it live in the terms and conditions?
- No. The notice must come at the start of the first interaction, clearly and distinguishably. Burying it in legalese doesn't comply.
- What if the agent only serves businesses?
- It still applies. The test is whether the counterpart is a natural person — consumer or professional.
- Who enforces it?
- Mainly national market surveillance authorities. The AI Office has a limited, narrowly scoped role.
Related terms
Sources & further reading
European Commission — Shaping Europe's digital future · 2026-07-24 · E3