AI-generated content: The EU publishes its transparency code

The European Commission published an AI Act code of practice to help providers and deployers comply with Article 50 transparency rules on generative AI.

The European Commission has made available a code of practice designed to facilitate compliance with the transparency obligations of the AI Act, in particular its Article 50, which governs generative AI systems.

These obligations, applicable from August 2, 2026, aim to limit the risks of deception and manipulation by requiring the labeling and detection of AI-generated content, as well as the reporting of deepfakes and certain publications. Drafted by independent experts in a process involving multiple stakeholders, under the auspices of the AI Office, the text is divided into two parts: one for providers, on labeling and detection, the other for deployers, on the labeling of deepfakes and generated or modified texts. A set of icons complements the framework.

Adherence remains voluntary, but the requirements of Article 50 remain legal obligations: signatories will be able to invoke the code to demonstrate their compliance and gain legal certainty across all Member States, whereas others will have to prove the adequacy of their own measures. Still under evaluation by the Commission and the AI Board, the code will be supplemented by guidelines specifying the scope of Article 50.