Commission Publishes Second Draft Code of Practice on AI Content Transparency

The European Commission has released the second draft of the Code of Practice on the transparency of AI‑generated content under the AI Act. Prepared by independent experts, the draft integrates written feedback from hundreds of stakeholders, including industry, academia, civil society, Member States via the AI Board, and Members of the European Parliament through the IMCO‑LIBE Working Group. Input was collected through an EU survey and dedicated meetings and workshops held in January 2026.

Compared to the first version, the new draft is more concise and flexible. It reduces compliance burden while improving legal clarity and technical feasibility. The text promotes the use of open standards for AI content marking and introduces the concept of a common EU icon for labelling, with the aim of simplifying compliance and lowering implementation costs for signatories.

Section 1 of the draft targets providers of generative AI systems falling under Article 50(2) of the AI Act. It introduces a revised two‑layer marking model based on secured metadata and watermarking, complemented by optional fingerprinting and logging. Several obligations from the earlier draft have been merged or made optional, reflecting a clearer focus on proportionate and technically workable measures.

Section 2 applies to deployers and addresses labelling obligations for deepfakes and certain text publications under Article 50(4). The revised section removes the distinction between AI‑generated and AI‑assisted content and replaces it with flexible design and placement requirements for labels, icons, or disclaimers. It also clarifies how artistic, satirical, fictional, and editorially controlled content may rely on existing practices. Feedback on the draft is open until 30 March 2026, with final adoption expected by early June, ahead of the transparency rules becoming applicable on 2 August 2026.