European Commission publishes Code of Practice on marking and labelling AI-generated content — Article 50 transparency operational guidance lands ahead of August 2 deadline
The European Commission published the final Code of Practice on marking and labelling AI-generated content on June 10, providing operational guidance for Article 50 transparency compliance. The Code lands roughly six weeks before the August 2, 2026 EU AI Act transparency deadline takes effect for general-purpose AI systems.
The Code is voluntary in form, mandatory in practice. EU enforcement guidance has been signaling that adherence to the Code of Practice will be treated as presumptive compliance with Article 50; deviation requires the deployer to demonstrate equivalent transparency by other means. For US-headquartered labs distributing models into the European market, that's a structured path to compliance — provided they engage now rather than wait for August 2.
The substantive piece is the labelling spec itself. The Code defines technical requirements for content provenance markers (think C2PA-style cryptographic signatures), user-facing disclosure language, and the granularity of "AI-generated" claims (full-output, edited, AI-assisted). For the deepfake category specifically, the Code is the first authoritative European reference framework. Combined with the CDT dark-patterns submission covered this morning, EU regulators now have two enforcement-ready inputs going into the August window.
European Commission — AI Act | Shaping Europe's digital future → · Global Policy Watch — EU AI Act Update: Timeline Relief, Targeted Simplification, and New Prohibitions →