Article 50 of the EU AI Act took effect on 2 August 2026, following the two-year transition period from when the regulation entered into force in August 2024. The article is now enforceable. National market surveillance authorities designated under Article 74 hold primary enforcement responsibility. The European Commission published non-binding guidelines on 20 July 2026 to clarify the scope of each obligation for competent authorities, providers, and deployers.
Article 50(1) requires providers of AI systems that interact directly with persons to design those systems so that users receive clear notification they are interacting with AI, unless the context makes this self-evident to a reasonably informed person. Article 50(3) requires providers of systems generating synthetic audio, image, video, or text content to mark outputs in a machine-readable format detectable as AI-generated. Article 50(4) requires deployers of deepfake systems to disclose the artificial origin of content. Non-compliance can attract fines of up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher.
The obligations fall on two distinct groups. Providers placing AI systems on the EU market must build disclosure mechanisms into the system before release. Deployers operating an AI system in a professional context must notify end-users prior to each interaction, except where the system is used to detect, prevent, investigate, or prosecute criminal offences. Operators of customer-facing chatbots, synthetic media platforms, AI-authored content services, and emotion-recognition tools face immediate compliance requirements as of 2 August 2026.
For generative AI systems already placed on the EU market before 2 August 2026, providers have until 2 December 2026 to achieve compliance with the machine-readable content-marking requirement. The Commission guidelines acknowledge that whether a human reviewer's edits qualify as a substantial alteration removing the AI-origin disclosure duty is an open question that national authorities will resolve case by case.
Licentium may advise on this matter or connect clients with our partner network. Work we undertake includes EU AI Act system classification, deployer obligation gap assessments, transparency notice design, watermarking and content provenance compliance, and regulatory correspondence with national market surveillance authorities.