Article 50 is now in force: who must do what
Last updated 1 August 2026
On 2 August 2026, the EU AI Act's transparency obligations — Article 50 of Regulation (EU) 2024/1689 — become applicable law. This is the Regulation's general application date under Article 113, and unlike the high-risk rules, it was not postponed by the 2026 Digital Omnibus (Regulation (EU) 2026/1744). Here is exactly what each paragraph requires, who it covers, and how to close the gap today.
Article 50(1) — tell people they are talking to an AI
Who: providers of AI systems intended to interact directly with natural persons — customer-support chatbots, AI sales assistants, voice agents, conversational search.
What: the system must be designed so that the people concerned are informed they are interacting with an AI system, unless this is obvious to a reasonably well-informed, observant and circumspect person given the circumstances and context of use.
How to comply: a clear notice at, or before, first interaction — in the chat window, in the voice greeting, in the interface. Our one-script disclosure badge renders this notice in 24 EU languages, and the chatbot disclosure examples guide shows wording that works.
Article 50(2) — mark synthetic content machine-readably
Who: providers of AI systems (including general-purpose systems) that generate synthetic audio, image, video or text content.
What: outputs must be marked in a machine-readable format and detectable as artificially generated or manipulated. Solutions must be effective, interoperable, robust and reliable as far as technically feasible.
How to comply: embed provenance metadata such as C2PA Content Credentials at generation time. Our C2PA signing API does this with one POST request — see the C2PA guide for the details.
Article 50(3) — disclose emotion recognition and biometric categorisation
Who: deployers of emotion recognition systems or biometric categorisation systems.
What: inform the natural persons exposed to the system of its operation, and process personal data in line with EU data protection law. (Note: certain law-enforcement uses are exempted.)
Article 50(4) — label deep fakes and AI public-interest text
Who: deployers of AI systems that generate or manipulate image, audio or video constituting a deep fake, and deployers publishing AI-generated or AI-manipulated text to inform the public on matters of public interest.
What: disclose that the content has been artificially generated or manipulated. Limited exemptions exist for artistic, creative, satirical or fictional works (disclosure in a way that does not hamper the display or enjoyment of the work) and where the text has undergone human review with editorial responsibility.
How to comply: a visible label on the published content — see how to label AI-generated content.
Article 50(5) — clarity and timing
All of the above must be provided in a clear and distinguishable manner at the latest at the time of the first interaction or exposure, and must conform to applicable accessibility requirements.
What non-compliance costs
Breaches of Article 50 obligations by providers and deployers carry administrative fines up to €15,000,000 or 3% of total worldwide annual turnover, whichever is higher (Article 99(4)(g)). The obligations reach non-EU companies whenever the system or its output is used in the EU (Article 2) — see what this means for non-EU companies.
Your 60-minute compliance path
- Run the free 3-minute checker to see which paragraphs apply to your product.
- Add the disclosure badge to every chatbot and AI-content surface.
- Mark generated media with the C2PA API.
- Export the dated evidence pack from the checker and file it. If you need policy documents, the template pack and plans cover the rest.
Sources
Not sure which obligations apply to you?
Run the free 3-minute Article 50 check