EU Mandates AI Content Labels by August 2 — Commission Publishes Watermarking Standards
Summary: The European Commission published its AI content labeling Code of Practice on June 10, setting technical requirements that become legally binding on August 2, 2026.
Key Facts
- Article 50 of the EU AI Act requires providers to embed machine-readable provenance metadata in all AI-generated or AI-manipulated content
- A standardized visual "AI" label is proposed, with a taxonomy distinguishing "fully AI-generated" from "AI-assisted" content
- Deployers using generative AI professionally must add explicit disclosures to deepfakes and AI-authored text on matters of public interest
- Technical standards cover watermarking, C2PA-compatible metadata embedding, and provenance tooling
- On June 1, the Commission also stood up a 60-member independent Scientific Panel to support AI Act enforcement across national authorities
Why It Matters
With about six weeks until the August 2 enforcement date, companies deploying generative AI to European users need to audit their watermarking and metadata pipelines now. Non-compliance with transparency obligations can draw penalties of up to 1.5% of global annual turnover — a significant exposure for any major AI provider. The Scientific Panel adds sustained technical oversight capacity, signaling the EU intends to enforce these rules rather than treat them as aspirational.
Read More
- Code of Practice on marking and labelling of AI-generated content — European Commission
- AI Act enforcement gets independent expert support — European Commission