The European Commission published final guidelines on the transparency obligations set out in Article 50 of Regulation (EU) 2024/1689, the EU AI Act. The guidelines were issued ahead of 2 August 2026, when Article 50 transparency obligations became binding on providers and deployers of the covered AI systems. This is final interpretive guidance, directly applicable in all EU member states without transposition.
Article 50(1) of Regulation (EU) 2024/1689 requires providers to design AI systems that interact directly with natural persons so those persons are informed they are interacting with an AI system at the point of interaction. Article 50(2) obliges deployers of emotion recognition systems and biometric categorisation systems to inform each natural person subject to those systems before processing begins. Article 50(4) requires providers and deployers of generative AI systems producing synthetic audio, video, image, or text to label that content in machine-readable form detectable by automated systems. The guidelines confirm that adherence to the Code of Practice on Transparency of AI-generated Content creates a presumption of conformity with Article 50(4) labelling duties, and that non-adherents must document alternative technically equivalent measures.
Providers placing interactive or conversational AI products on the EU market must build Article 50(1) disclosure into the system design before first deployment. Deployers running emotion recognition systems in employment screening, retail analytics, or access control must implement user-facing notification at the point of processing. Broadcasters, media platforms, and advertising networks distributing synthetic content must integrate machine-readable watermarking or metadata embedding capable of automated detection. Providers and deployers outside the Code of Practice must document their alternative technical measures and be prepared to demonstrate equivalency to national market surveillance authorities.
Article 50(1) disclosure is not required where the AI nature of the interaction is obvious from context to a reasonably well-informed user. An exception under Article 50(4) permits providers to omit labelling in creative, artistic, or clearly fictional contexts where the AI-generated nature is explicitly acknowledged. The guidelines clarify that the Code of Practice adequacy decision covers only Article 50(4) marking obligations and does not create compliance coverage for Articles 50(1) or 50(2). Enforcement authority rests with national market surveillance authorities designated under Article 70 of the AI Act.
We advise AI system providers and deployers on compliance obligations under Regulation (EU) 2024/1689 and may assist through our partner network across EU member state jurisdictions. Contact us to assess your Article 50 position, Code of Practice adherence strategy, or supervisory authority exposure. Work we undertake includes AI Act compliance gap assessments, Article 50 transparency obligation mapping, Code of Practice adherence analysis, and regulatory advisory for AI providers and deployers in the EU.