The EU Code of Practice on AI content marking — six weeks before August 2, the labelling spec gets concrete
The Commission's June 10 Code of Practice on marking and labelling AI-generated content is the first operational deliverable inside the Digital Omnibus on AI package. For frontier labs preparing transparency disclosures by August 2, the Code is the structured compliance path.
The European Commission's Code of Practice on marking and labelling AI-generated content dropped Tuesday with six weeks to go before the AI Act's August 2 transparency obligations bind general-purpose AI providers. The Code is voluntary in form but operationally mandatory: adherence is presumptive compliance with Article 50, and deviation requires the deployer to demonstrate equivalent transparency another way.
What the labelling spec actually covers
The Code defines technical requirements for content provenance markers (C2PA-style cryptographic signatures), user-facing disclosure language (the specific phrasing for "AI-generated," "AI-edited," "AI-assisted" labels), and the granularity of provenance claims (full-output vs. partial editing vs. assisted authoring). For the deepfake category specifically, the Code is the first authoritative European reference framework that names the technical and disclosure requirements.
The provenance-supply-chain layer becomes procurement-relevant
Microsoft's MAI-Thinking-1 zero-distillation disclosure already treats provenance as a model-supply-chain marketing axis. The Code extends that into the content-output supply chain: not just "what's in the model" but "how was this output created." Enterprise procurement will increasingly ask both questions simultaneously, which means content-generation vendors need provenance tooling that spans both the model layer and the output layer.
How the Digital Omnibus reshapes the timing
The Digital Omnibus on AI heads for final adoption in June with publication in July. The amendments push most high-risk-system obligations into 2027 and clarify the boundary between high-risk and general-purpose obligations. Combined with the Code of Practice, the EU has now published both the operational guidance (Code) and the legislative framework (Omnibus) that govern August 2 compliance for the first time.
What US labs do now
For OpenAI, Anthropic, Google, and Microsoft distributing models into the European market, the next six weeks are the engagement window. Provenance tooling has to be production-ready by August 2 if the lab wants Code-of-Practice compliance status. Trump's June 2 EO formalizes a different US-side disclosure regime that doesn't conflict with the EU framework — meaning labs can run both compliance programs without contradiction. For the cross-jurisdictional procurement story, that's a cleaner regulatory map than the labs have had at any point in 2026.
European Commission — AI Act | Shaping Europe's digital future → · Global Policy Watch — EU AI Act Update →