The European Commission is preparing the code for labeling content generated by artificial intelligence.

The European Commission's Office for Artificial Intelligence has organized a new round of meetings and workshops to collect feedback on the future Code of Practice regarding the marking and labeling of AI-generated content, ahead of the publication of the third and final draft, expected at the beginning of June. In short The code addresses the transparency obligations in Article 50 of the AI Act, which concern content generated or manipulated by artificial intelligence systems.

The discussions covered the marking of AI content in machine-readable format, the detection of synthetic content, the labeling of deepfakes, and informing the public when AI-generated texts are used on topics of public interest.

The industry called for flexibility, legal clarity, and the avoidance of technically difficult obligations, while civil society organizations and the academic environment insisted on clear, accessible, and effective labels.

The transparency rules from the AI Act are set to become applicable on August 2, 2026, and the code is expected to function as a voluntary compliance tool for providers and professional users of AI systems. The meetings took place in March 2026 and involved two working groups. The first group focused on providers of generative AI systems and their obligation to make AI-generated content detectable, through solutions such as metadata, watermarking, verification mechanisms, or other forms of technical marking. The discussions in this group addressed a multi-level approach to content marking, the role of providers in ensuring detectability, technical feasibility, performance assessment of solutions, possible compliance mechanisms, and the use of evaluations conducted by third parties. According to the summary published by the Commission, participants had differing views on how prescriptive the rules should be. Industry representatives cited the burden of compliance, costs for innovation, and the technical limitations of some solutions. Civil society and academic representatives called for strong guarantees, so that the future code protects the public interest and citizens' right to know when interacting with artificially generated or manipulated materials. The second working group focused on professional users of AI systems, particularly the obligations to signal deepfakes and certain AI-generated texts published on topics of public interest. Discussions included how user information should be formulated, the placement of labels, proportionality of requirements, editorial control, and the development of a uniform European label. Participants also discussed the risk of "fatigue" among users regarding labels, possible areas of legal safety for actors applying the rules in good faith, and the need for coherence between the AI Act, Digital Services Act, and other applicable European rules for online platforms. Alongside the meetings of the two working groups, the Commission organized three additional workshops. One analyzed the design, placement, and testing of labels for deepfakes and AI-generated texts. Another focused on the technical details of content marking, including public or secure metadata, watermarking, and possible alternatives to watermarking. A joint workshop discussed how machine-readable markings, provenance data, visible labels, and a possible common European symbol could work together. The Commission states that the feedback received in these meetings, along with additional written contributions, will be used to draft the third and final version of the code. The AI Act establishes transparency obligations for providers and professional users of certain AI systems. In the case of artificially generated or manipulated content, the goal is for the public to more easily identify synthetic materials, including deepfakes or AI-generated texts used to inform the public on topics of public interest. The code of practice does not replace the law, but if approved by the Commission, it could be used voluntarily by companies to demonstrate compliance with the obligations set out in the AI Act.