Code of Practice on Transparency of AI-Generated Content
On 10 June 2026, the European Commission published the final Code of Practice on Transparency of AI-Generated Content, a voluntary instrument designed to help providers and deployers of generative AI systems comply with the transparency obligations established by the EU AI Act. These obligations will start to apply on 2 August 2026.
The Code addresses one of the most pressing challenges raised by generative AI: how to ensure that citizens can recognise when content has been created or manipulated by artificial intelligence. As synthetic text, images, audio and video become increasingly sophisticated, transparency is emerging as a key tool for preserving trust in the information ecosystem and reducing the risks of misinformation, manipulation and deception.
The document is divided into two sections:
- The first concerns providers of generative AI systems and explains how AI-generated or AI-manipulated content should be marked and made detectable. The Code is structured around four core commitments concerning the marking of AI-generated content, the provision of detection mechanisms, compliance with quality requirements such as effectiveness, reliability, robustness and interoperability, and the establishment of testing and compliance procedures. To this end, it promotes a multi-layered approach based on machine-readable metadata, watermarking techniques and detection mechanisms capable of identifying synthetic content.
- The second section focuses on deployers of AI systems, clarifying their obligations to label deepfakes and AI-generated or AI-manipulated content published on matters of public interest. The objective is to guarantee that users receive clear information when interacting with synthetic content and can make informed decisions about its reliability.
Although voluntary, the Code is likely to play an important role in the implementation of the AI Act. Once approved by the European Commission and the AI Board, adherence to the Code may help organisations demonstrate compliance with the transparency requirements set out in Article 50 AI Act.
Beyond its practical relevance, the initiative reflects the broader European approach to AI governance: promoting innovation while ensuring accountability, transparency and the protection of fundamental rights. In this respect, the Code represents another significant step towards the development of a trustworthy and human-centric AI ecosystem in Europe.
The publication of the Code does not conclude the implementation process. The European Commission has announced that the Code will be complemented by forthcoming Guidelines on the transparency obligations under Article 50 of the AI Act. These Guidelines are expected to clarify the scope of the legal requirements, address issues not covered by the Code, and provide practical guidance for providers and deployers. Their publication will be particularly important in resolving interpretative questions concerning the notions of AI-generated and AI-manipulated content, the disclosure obligations applicable to deepfakes, and the relationship between transparency requirements and editorial control. As such, the Code should be viewed as an important first step within a broader regulatory framework that is still evolving.
