The European Commission published the Code of Practice on Transparency of AI-Generated Content in June 2026. The Code implements the mandatory transparency obligations under Articles 50(2) and (4) of Regulation (EU) 2024/1689 (the AI Act), which take effect on 2 August 2026. The Code is published in final form. It is voluntary in adoption but translates legally binding AI Act obligations into practical implementation guidance; a firm may meet Article 50 requirements by following the Code or by a demonstrably equivalent alternative approach.
Article 50(2) requires providers of AI systems that generate or manipulate image, audio, or video content to ensure their outputs carry machine-readable marks at the time of generation, enabling automated detection of synthetic origin. Article 50(4) requires deployers who use AI systems to produce or modify video, image, or audio content constituting a deepfake to disclose to persons exposed to that content that it was artificially generated or manipulated. Where the content concerns matters of public interest, including news and election-related material, Article 50(3) extends the disclosure obligation to AI-generated text. The Code provides an optional EU icon in three variants that deployers may use to satisfy the labelling obligation in a standardised way.
The compliance obligations fall on two distinct categories of firm. Providers of generative AI models, including image generators, voice synthesis tools, video generation platforms, and multimodal models with image or audio output, must implement technical watermarking or equivalent marking mechanisms before their systems are placed on the EU market after 2 August 2026. Deployers operating consumer-facing products that present AI-generated or AI-manipulated content to end users must implement disclosure workflows at the point of content exposure. This affects media platforms, fintech and crypto marketing teams using AI-generated imagery or voiceovers, and any product generating AI avatars, AI-edited video, or AI-synthesised promotional material.
The Code provides carveouts for artistic, creative, satirical, or fictional works where human editorial review and responsibility are exercised. These carveouts are narrow; they apply only where the creative or fictional nature of the content is apparent from context, and deepfake elements in non-satirical contexts are unlikely to qualify. National supervisory authorities retain discretion on what constitutes adequate labelling. Article 50 obligations run in parallel to member state rules on synthetic media and to Digital Services Act obligations on platforms concerning AI-generated content.
Licentium advises AI providers, media companies, fintech firms, and digital product teams on compliance with Article 50 transparency obligations under the EU AI Act. If your product generates or distributes synthetic media reaching EU users, we can assess your obligations and help implement compliant disclosure workflows before 2 August 2026. Work we undertake includes AI Act compliance reviews, deepfake disclosure analysis, technical marking requirement assessments, and carveout applicability opinions.
Source: European Commission, Code of Practice on Transparency of AI-Generated Content, June 2026