The European Commission published the final Code of Practice on Transparency of AI-Generated Content on 10 June 2026 via the digital-strategy.ec.europa.eu platform. The Code is the implementation instrument for Article 50 of EU Regulation 2024/1689, which requires providers and deployers of AI systems that generate or manipulate audio, image, text, or video to mark and disclose AI-generated outputs. Article 50 enters into application on 2 August 2026, the same date as the broader general-purpose AI model transparency obligations under Article 53.
Article 50(1) of EU Regulation 2024/1689 requires providers of AI systems with generation or manipulation capabilities to mark outputs in a machine-readable format. Article 50(4) requires deployers who use such systems to generate content intended for the public to disclose that the content is AI-generated, unless the disclosure would be disproportionate or the AI character is evident from context. The Code specifies that conforming implementations must use digitally signed metadata and imperceptible watermarking meeting C2PA or equivalent provenance standards, and proposes a harmonised EU content provenance label. Adherence to the Code creates a presumption of conformity under Article 56 of the AI Act for the obligations it covers.
Providers of general-purpose AI models with generation capabilities must implement machine-readable marking at the model output layer. Deployers making AI-generated content publicly available must add user-facing disclosure at the point of delivery. Financial institutions using AI to generate client-facing documents, reports, or marketing materials must comply with the Article 50 disclosure obligation for any output that could influence an investor, borrower, or consumer decision. Organisations that submit adherence forms by 22 July 2026 will be listed as initial signatories and benefit from the presumption of conformity in any market surveillance inquiry from 2 August 2026.
The disclosure obligation does not apply where the AI character of the content is evident from context, or where the content undergoes proportionate human editorial review making the operator responsible for its accuracy. Open questions include the definition of proportionate human review sufficient to displace the disclosure duty, the scope of the obligation for AI-assisted rather than fully AI-generated content, and the harmonised technical watermarking standards, which CEN/CENELEC is expected to finalise before end-2026. The Code is voluntary, but non-signatories must demonstrate compliance with Article 50 through alternative means.
Licentium advises technology companies, financial institutions, and media operators on EU AI Act compliance, including Article 50 transparency obligations and Code of Practice adherence. We assist with disclosure programme design, European AI Office engagement, and compliance documentation across EU Member States. Work we undertake includes AI Act transparency audit, Code of Practice signature and adherence review, deployer liability assessment, and AI-generated content governance.
Source: European Commission, Code of Practice on Transparency of AI-Generated Content (Final), 10 June 2026