Article 50(2) of the EU AI Act requires providers of generative AI to mark their outputs — synthetic audio, image, video and text — in a format that is machine-readable and detectable as artificially generated. To help the market implement that duty consistently, the European Commission has published a Code of Practice on the transparency of AI-generated content.
The Code translates a short legal sentence into practical engineering expectations: what "machine-readable" marking means in practice, how detection should work across content types, and how deployers of deepfakes and public-interest AI-generated text should surface the disclosure to users. It is aimed at closing the gap between the obligation and the patchwork of watermarking and provenance techniques currently in use.
Timing makes it urgent. Article 50 transparency obligations apply from 2 August 2026, with a four-month grace period until 2 December 2026 for the synthetic-content marking obligation on systems already on the market before that date. The Code gives providers and deployers a reference point to align on before enforcement begins, rather than discovering after the fact that their labelling is not detectable in the way the Act intends.
For organisations, the immediate action is to confirm that any generative tooling they ship or embed emits conformant, detectable marking — and to keep the vendor attestations as evidence. Veritome links those attestations to the Article 50 obligation on each affected system.
Sources: European Commission, "Code of Practice on AI-generated content"; Tech Policy Press, "The EU's AI Transparency Code of Practice, Explained".