EU publishes code for labeling AI-generated content

The European Commission published a voluntary Code of Practice on marking and labeling AI-generated content on June 10, 2026, following a public consultation. The Code gives providers and deployers a pathway to demonstrate compliance with EU AI Act transparency duties, creating a presumption of compliance for signatories while leaving companies that choose alternative methods responsible for proving equivalent safeguards.

Providers of generative AI systems are expected to use machine-readable markings for AI-generated audio, image, video, and text. The Code says most signatories will need a multi-layered approach, combining metadata where supported with imperceptible watermarks, while very short text under 200 tokens is exempt from watermarking. Providers must also offer detection tools that are clear, generally free of charge, privacy-preserving, and reliable across content types and common manipulation techniques.

Deployers must disclose deepfakes and AI-generated or manipulated text on matters of public interest, using a public EU icon or equivalent labels that are immediately perceivable. Labels must appear prominently in formats such as videos, published text, and audio outputs. Transparency obligations begin on August 2, 2026, while providers face additional compliance milestones on December 2, 2026, and February 2, 2027.