The European Commission published the Draft Guidelines on the Implementation of the Transparency Obligations of Certain AI Systems Under Article 50 of the AI Act on 8 May 2026. The guidelines clarify Commission expectations under Article 50 of Regulation (EU) 2024/1689. The instrument is a draft. A targeted consultation closes on 3 June 2026. The Article 50 obligations bind providers and deployers from 2 August 2026.
Article 50(1) requires providers of AI systems that interact directly with people to design those systems so users know they deal with an AI system. Article 50(2) requires providers of generative AI systems to mark outputs as artificially generated or manipulated in a machine-readable form. Article 50(3) requires deployers of emotion recognition or biometric categorisation systems to inform people exposed to the system. Article 50(4) requires deployers of deepfake systems to label artificially generated or manipulated image, audio, or video content. The draft guidelines explain scope, exemptions, and the interaction between provider and deployer duties.
Providers of consumer chatbots, virtual assistants, customer service agents, and conversational interfaces face concrete disclosure-by-design requirements. Image, video, audio, and text generation services owe machine-readable provenance marks on outputs. Workplace, retail, education, and security operators using emotion recognition or biometric categorisation owe individual notice to exposed persons. Media platforms, news organisations, advertising buyers, and political campaign teams deploying deepfake content owe labelling at the user-visible layer. Vendors selling generative AI to EU clients must give those clients the compliance hooks the deployer duties require.
Article 50(4) labelling rules carve out content that forms part of a manifestly artistic, creative, satirical, or analogous work. Article 50(2) marking rules exempt outputs from standard editing functions and outputs that do not substantially alter the input. Law enforcement uses sit under a separate carve out. The draft guidelines are non-binding. The technical detail for Article 50(2) and Article 50(4) sits in a separate Code of Practice, expected in final form in June 2026. The Article 50 obligations apply from 2 August 2026.
Licentium advises crypto, AI, and digital asset clients on EU AI Act compliance, with a partner network for member state local counsel. Work we undertake includes Article 50 mapping, transparency UX reviews, generative AI provenance design, deepfake labelling policy, deployer notice drafting, and engagement with the Commission and national competent authorities. Contact us to align your Article 50 build with the draft guidelines before 2 August 2026.
Source: European Commission, Draft Guidelines on the Implementation of the Transparency Obligations of Certain AI Systems Under Article 50 of the AI Act, published 8 May 2026, https://digital-strategy.ec.europa.eu/en/library/draft-guidelines-implementation-transparency-obligations-certain-ai-systems-under-article-50-ai-act
The information provided is not legal, tax, investment, or accounting advice and should not be used as such. It is for discussion purposes only. Seek guidance from your own legal counsel and advisors on any matters. The views presented are those of the author and not any other individual or organization. Some parts of the text may be automatically generated. The author of this material makes no guarantees or warranties about the accuracy or completeness of the information.