The European Commission's AI Office published the final Code of Practice on marking and labelling of AI-generated content on 9 June 2026, following a public consultation on an earlier draft. The Code is voluntary and establishes the technical measures and transparency practices through which providers and deployers of generative AI systems can demonstrate compliance with the mandatory transparency obligations under Article 50 of Regulation (EU) 2024/1689 (the EU AI Act), which take effect on 2 August 2026.
The Code is structured in two sections. Section 1 addresses provenance requirements for providers of generative AI systems, prescribing digitally-signed metadata embedding (sub-measure 1.1.1) and imperceptible watermarking (sub-measure 1.1.2) as the primary technical mechanisms for marking AI-generated content at the point of generation; fingerprinting or logging via a registry database (sub-measure 1.1.3) is an optional supplementary mechanism. Section 2 addresses transparency obligations for deployers, requiring visible labels on deepfakes and AI-generated or AI-manipulated text published on matters of public interest, and mandatory user disclosure whenever a person interacts with an AI chatbot or virtual assistant, irrespective of the subject matter of the interaction. The legal obligation anchored in Article 50 AI Act applies to providers of general-purpose AI models and deployers of AI systems generating synthetic content distributed to end users.
Providers of large language models, image-generation systems, audio synthesis tools, and video-generation models must implement provenance metadata embedding before 2 August 2026 or document a reasoned equivalent technical approach for national market surveillance authorities. Deployers — including media platforms, marketing automation tools, customer service chatbot operators, and social media platforms distributing AI-generated content — must add visible AI-origin labels to synthetic text, images, audio, and video published on matters of public interest. All operators of AI chatbots and virtual assistants accessible to EU users must display a clear disclosure at the point of user interaction.
The Code does not carry the binding force of the AI Act itself; divergence from Code measures is permitted where an operator can demonstrate to national market surveillance authorities that an alternative approach satisfies Article 50 with equivalent effect. The AI Office has indicated that sectoral guidance for specific deployment contexts — including AI in online platforms, journalism, and advertising — is expected later in 2026. The threshold for what constitutes a matter of public interest, triggering the visible labelling obligation for deployers, is not defined in the Code and may be addressed in subsequent guidance.
Licentium may advise clients on EU AI Act compliance and has a partner network to assist AI providers, deployers, and distributors preparing for the August 2026 transparency obligations. Companies assessing their Article 50 implementation, reviewing content labelling pipelines, or calibrating chatbot disclosure obligations are welcome to contact us. Work we undertake includes EU AI Act compliance analysis, Article 50 technical specification review, AI system classification, transparency notice drafting, and AI regulatory strategy.
Source: European Commission AI Office, Code of Practice on Transparency of AI-Generated Content, 9 June 2026