The European Commission is expected to publish the final version of the Code of Practice on marking and labelling AI-generated content in June 2026, following a drafting process that has involved multiple working group sessions, two published drafts, and a stakeholder consultation running since late 2025. The Code provides the practical framework for compliance with Article 50 of the EU AI Act, whose transparency obligations become applicable on 2 August 2026.
Article 50 imposes two distinct sets of obligations. Providers of AI systems that generate synthetic audio, image, video, or text content must ensure their outputs are marked in a machine-readable format that allows downstream detection. This applies to all generative AI systems, regardless of whether the output is intended to be realistic or clearly artificial. Deployers - organisations using these systems for professional purposes - must disclose when content constitutes a deepfake (an image, audio, or video that resembles real persons, places, or events and could falsely appear authentic) and must clearly label AI-generated text when it is published on matters of public interest.
The Code of Practice is structured in two sections. Section 1 addresses marking and detection, aimed at providers of generative AI systems within the scope of Article 50(2). Section 2 targets deployers, focusing on the labelling of deepfakes and AI-generated text publications. The second draft, published on 5 March 2026, refined the technical requirements for machine-readable marking and introduced guidance on when labelling exceptions apply - for example, where AI-assisted editing does not substantially alter the original content.
The Commission published the Code alongside draft Guidelines on Article 50 for stakeholder consultation, providing interpretive guidance on scope, definitions, exceptions, and horizontal issues. Both instruments are designed to be read together, giving providers and deployers a practical compliance pathway ahead of the August deadline.
Acompli perspective: August 2026 is two months away, and the transparency obligations apply to any organisation that deploys generative AI systems professionally - not just the providers that build them. If your organisation uses AI to generate content that could be published, shared externally, or mistaken for human-created material, you need to understand whether you fall within the deployer obligations under Article 50. Reviewing your risk assessments to cover AI-generated content, and ensuring your records of processing reflect where generative AI systems are used and what outputs they produce, are practical first steps.