EU finalizes AI transparency code before Article 50 deadline

The European AI Office’s finalized Code of Practice on Transparency of AI-Generated Content sets out how providers and deployers in the EU should approach Article 50 transparency duties. The rules apply to providers of systems that generate synthetic audio, image, video or text, and to deployers using AI for deepfakes or text published to inform the public on matters of public interest.

The final code makes several changes from the first draft published in December 2025. Some measures are now optional, including provenance transparency, perceptible marking functionality and forensic detection mechanisms. The code also expands marking flexibility, introduces a finalized EU icon using the acronym AI, and sets an interoperability deadline of 2 February 2027 for watermark detection mechanisms.

Signatories must implement compliance processes proportionate to their size, while deployers must disclose deepfakes and certain AI-generated public-interest text using the EU icon or an equivalent label. Financial market participants using generative AI for public communications such as macroeconomic outlooks, sector reports, market commentary or environmental, social, and governance reports are encouraged to assess whether Article 50 obligations are triggered and whether adherence to the code would support regulatory compliance.