On 20 July 2026, the European Commission published guidelines on the transparency obligations under Article 50 of Regulation (EU) 2024/1689. The guidelines are non-binding interpretation documents issued ahead of the 2 August 2026 application date. Their purpose is to assist providers and deployers of AI systems in determining which obligations apply to their systems and how to meet them.
The guidelines interpret Article 50, paragraphs 1 through 4. On Article 50(1), the document defines when an AI system interacts directly with natural persons and when the artificial nature is obvious from context, thereby triggering or excluding the disclosure obligation. On Article 50(2), the guidelines address what constitutes a generative AI system and which technical methods satisfy the machine-readable marking requirement. On Article 50(4), the document defines matters of public interest and identifies the conditions under which the absence of human review activates the labelling obligation.
Providers of AI systems across media, financial services, healthcare, and public administration benefit most directly from the scope clarifications. AI system providers need the guidelines' interpretation of machine-readable marking technical standards to ensure their systems support downstream deployer compliance. Deployers publishing or distributing AI-generated text rely on the guidelines' definition of public interest content to determine whether the Article 50(4) labelling duty applies to their outputs.
The guidelines carry no binding legal force. National market surveillance authorities designated under Article 74 of the AI Act may depart from the Commission's readings in enforcement proceedings. The Commission also published a voluntary Code of Practice on Marking and Labelling of AI-generated Content; the guidelines treat that Code as complementary to, not a substitute for, the Article 50(2) statutory obligation. Open questions on the interaction between Article 50's requirements and disclosure duties under Regulation (EU) 2016/679 and Directive 2005/29/EC remain unresolved pending guidance from competent national authorities.
Licentium advises on EU AI Act compliance, including scope analysis and application of Article 50 for providers and deployers. Work we undertake includes Article 50 obligation assessments, provider and deployer role classification under Regulation (EU) 2024/1689, machine-readable marking strategy, and coordination with parallel GDPR and consumer protection obligations.