From the journal

EU AI Act Article 50 transparency obligations enter application on 2 August 2026

Article 50 of Regulation (EU) 2024/1689 takes effect across all EU member states on 2 August 2026, requiring providers and deployers of AI systems to disclose AI interactions, label AI-generated content, and implement machine-readable detection marks in generative AI outputs. The obligations cover providers of general-purpose AI models, deployers of emotion recognition and biometric categorisation systems, and entities producing AI-generated content on matters of public interest.

3 min read

Article 50 of Regulation (EU) 2024/1689 (the EU AI Act) enters into application on 2 August 2026, marking the transparency phase of the Act's phased roll-in. The EU AI Act entered into force on 1 August 2024. Prohibited AI practices took effect on 2 February 2025. General-purpose AI model obligations entered application on 2 August 2025. Article 50, alongside most high-risk AI system requirements, takes effect on 2 August 2026.

Article 50 imposes four disclosure duties. First, providers of AI systems designed to interact directly with people must inform users they are communicating with AI, unless the AI nature is evident from context. Second, providers of generative AI systems must implement machine-readable marks in outputs enabling detection of synthetic content, referencing technical standards under Article 50(7). Third, deployers using biometric categorisation, emotion recognition, or deepfake AI systems must disclose that use to affected persons. Fourth, any person producing AI-generated content on matters of public interest must label that content as AI-generated or AI-manipulated.

The obligations reach providers of large language models, image generators, synthetic audio tools, and other generative AI systems placed on the EU market. Deployers using third-party AI in their products or services bear separate disclosure duties toward end users. Media operators, political advertisers, professional service firms producing AI-generated client-facing documents, and operators of customer-facing AI chatbots must each assess their Article 50 obligations. Machine-readable marks must be embedded in AI output pipelines from 2 August 2026.

Each EU member state designates a competent authority and market surveillance authority under Article 70 to enforce Article 50. Fines fall under the general AI Act penalty regime, reaching up to EUR 15 million or 3% of global annual turnover for relevant violations. The European Commission published a Code of Practice on marking and labelling of AI-generated content in 2026 to guide compliance, but adherence does not substitute for meeting the statutory requirements. GPAI model providers face additional transparency obligations under Article 53, which entered application on 2 August 2025.

Licentium advises providers and deployers of AI systems on EU AI Act compliance, covering Article 50 disclosure obligations, GPAI model governance, and national implementation requirements across EU jurisdictions. We maintain a partner network with EU regulatory counsel for AI Act conformity assessment and enforcement readiness. Work we undertake includes EU AI Act compliance gap analysis, technical disclosure programme design, GPAI model documentation, and multi-jurisdictional AI regulatory mapping.

Source: European Commission, Regulation (EU) 2024/1689 (EU AI Act), Article 50, applicable from 2 August 2026

AI Regulatory

More from the journal

See all

Illinois signs Artificial Intelligence Safety Measures Act into law, effective January 2027

Governor JB Pritzker signed the Illinois Artificial Intelligence Safety Measures Act on 6 July 2026, making Illinois the first US state to require mandatory annual third-party audits of frontier AI model developers. The Act applies to large frontier developers operating or deploying models in Illinois, requires transparency reports before each new or substantially modified frontier model deployment, and establishes civil penalties. Core obligations take effect 1 January 2027.

UK consults on modernising payment services regulation covering stablecoins and AI agents, 2026

HM Treasury launched a consultation on modernising the UK payment services regulatory regime, with responses due by 6 October 2026. The consultation addresses how to adapt regulation for tokenised payments including qualifying stablecoins, Open Banking expansion, and agentic AI systems authorised to conduct payments. It follows the February 2026 Payments Forward Plan and the government's decision to transfer Payment Systems Regulator functions to the Financial Conduct Authority.

Hong Kong SFC mandates phishing-resistant authentication for internet brokers and VATPs, July 2026

The Hong Kong Securities and Futures Commission issued a circular on 9 July 2026 requiring licensed corporations engaged in electronic trading and SFC-licensed virtual asset trading platform operators to replace one-time passwords with phishing-resistant authentication methods. Covered firms must implement passkeys or bound-device recognition for client login and device binding. Large internet brokers must act immediately. All other in-scope firms have up to 12 months to comply.