The European Commission AI Office has completed a third round of meetings and workshops linked to the future Code of Practice on marking and labelling AI-generated content. The discussions will inform the final draft of the code, expected in early June.
The process forms part of implementation work under Article 50 of the Artificial Intelligence Act, which introduces transparency obligations for certain generative AI systems, deepfakes, and AI-generated content.
The Code of Practice is intended to clarify how providers and deployers of generative AI systems should identify, mark, detect, and disclose AI-generated material in practice.
Two working groups
Two separate working groups examined different parts of the problem. One focused on technical marking and detection obligations for providers of generative AI systems. The other focused on disclosure obligations for deployers using AI-generated text or deepfakes.
Stakeholders across the AI value chain participating in the code of practice process include providers of generative AI systems and models, technology providers of marking and detection tools, industry representatives, civil society organisations, and academic experts.
The discussions revealed significant disagreements between industry representatives, civil society groups, and academic experts. Meeting notes for the Third Round are linked here and below.
Industry participants raised concerns about compliance burdens, technical feasibility, innovation constraints, and risks linked to overly prescriptive rules. Civil society organisations and researchers argued for stronger safeguards, clearer disclosures, and more consistent transparency requirements across platforms.
Technical issues
Several technical issues were debated during the workshops. These included visible labels for AI-generated content, machine-readable watermarks, metadata systems, provenance information, deepfake disclosure mechanisms, and the possible introduction of a common EU transparency icon.
The practical challenge is interoperability across the AI content chain. AI-generated material can move between platforms, formats, and services, making it difficult to ensure that labels, watermarks, or provenance signals remain visible and technically reliable after editing, reposting, or transformation.
Another issue is proportionality. Smaller providers and deployers warned that complex watermarking, detection, and compliance systems could create operational burdens that are easier for large platforms to absorb than for SMEs or smaller developers.
Why it matters
The EU’s approach to artificial intelligence is intended to “promote excellence and trust, by boosting research and industrial capacity while ensuring safety and fundamental rights.”
The Code of Practice aims to support compliance with the AI Act transparency obligations under Article 50 of the AI Act (transparency obligations for providers and deployers of generative AI systems), which are intended to address risks of deception and manipulation, fostering the integrity of the information ecosystem. These transparency obligations pertain to marking and detection of AI generated content and labeling of deep fakes and certain AI generated publications.
The code itself will not replace the AI Act, but it is expected to shape how transparency obligations are interpreted operationally across the EU once enforcement begins.
Further reading
EU AI transparency code enters final drafting stage amid disputes over deepfake labels and watermarking. Press release. May 25, 2026. Civil Society Alliances for Digital Empowerment (CADE)
AI Act transparency code of practice – Third round of working group meetings. Article. May 22, 2026. European Commission
Code of Practice on marking and labelling of AI-generated content. Landing page. Accessed June 2, 2026. European Commission
European approach to artificial intelligence. Landing page. Accessed June 2, 2026. European Commission
AI Act. Policy framework (Landing page). Accessed June 2, 2026. European Commission
