European Commission publishes Code of Practice on Transparency of AI-Generated Content

IRIS 2026-6:1/19

Eric Munch

European Audiovisual Observatory

On 10 June 2026, the European Commission published the Code of Practice on Transparency of AI-Generated Content. The code, drawn up by independent experts, sets out practical steps to comply with the obligations stemming from the AI Act’s Article 50, on transparency with regard to the marking and labelling of AI-generated content.

As indicated in the code itself, its overarching objective is to:

“improve the functioning of the internal market, to promote the uptake of human-centric and trustworthy artificial intelligence and to support innovation pursuant to Article 1(1) AI Act, while ensuring a high level of protection of health, safety, and fundamental rights enshrined in the Charter, including democracy, the rule of law, and environmental protection, against harmful effects of AI in the Union.”

The code is divided into two sections: section 1 on rules for the marking and detection of AI-generated and manipulated content applicable to providers of generative AI systems (Article 51(2) and (5) AI Act) and section 2 on rules for labelling deep fakes and AI-generated and manipulated published text applicable to deployers of AI systems (Article 50(4) and (5) AI Act).

Signatories to the code agree to a series of commitments:

Under section 1, with regard to the marking of AI-generated or manipulated content, they commit to implementing marking solutions and to making them non-removable to the best of their abilities. Signatories are also encouraged to go further than what is required with regard to the information contained in the metadata indicating whether the content is AI-generated or manipulated. Providers of AI systems capable of generating or manipulating content are also encouraged to provide an optional functionality in their interface allowing deployers and other users to directly apply a perceptible label to the content generated or manipulated.

Signatories commit to making available a detection solution to allow for the verification of whether content has been AI-generated or manipulated. Detection results will have to be presented in a way that is clear and easily comprehensible to natural persons. Additionally, they may include a forensic detection mechanism to detect such content, complying with privacy and security requirements. They are also encouraged to ensure that relevant information is provided to deployers and expert users to support them in making informed decisions on what marking and detection solutions they may use (including helping them to understand how to access detection solutions, how to perform detections, and how to interpret detection results). They are also encouraged to ensure that end-user literacy resources are provided.

In addition, they commit to implementing effective solutions, with effectiveness being conditioned on the results of the detections being understandable by natural persons. The solutions must also be reliable (able to accurately assess when content was AI-generated or manipulated), robust (able to perform well under various conditions) and interoperable (able to operate seamlessly across multiple systems, actors, contexts and technical implementations). Signatories are also encouraged to invest in scientific research and development and to cooperate with competent authorities and relevant stakeholders.

Signatories commit to documenting, implementing, and keeping up to date a compliance process that describes how they have implemented the different measures. They also commit to testing the compliance of their marking and detection solutions with the code prior to placing their generative AI system on the market. Signatories also commit to providing their personnel who have roles relevant to ensuring compliance with Article 50(2) and (5) and Article 4 of the AI Act with proper training. Finally, they also commit to cooperating with the market surveillance authorities under the AI Act.

Under section 2, signatories commit to implementing specific design specifications (use of icons or labels, possibly accompanied by an acronym) positioned so as to be immediately recognised and noticed under normal conditions, and directly embedded into the content (unless user interface overlay is possible). Audible disclosure can be used as an additional method. Signatories are encouraged to take part in a dedicated task force aimed at advancing the further development and usability of the EU icon as a minimum state-of-the-art implementation.

Signatories commit to putting in place or maintaining internal processes, awareness measures and review mechanisms as specified in the measures below and proportionate to their size and available resources.

In accordance with Article 50(4) AI Act, they also commit to implementing measures to disclose deep fakes that are part of “evidently artistic, creative, satirical, fictional or analogous work or programmes” in a way that does not hamper the display or enjoyment of the work, including its normal exploitation and use, while maintaining the utility and quality of the work.

Lastly, all signatories commit to establishing, adapting, or maintaining appropriate policies for human review or editorial control prior to publication and that a natural or legal person holds editorial responsibility for the publication. Signatories who are media service providers under Article 2(2) of the European Media Freedom Act (EMFA) may rely on the exception to the disclosure obligation in Article 50(4), subparagraph 2, AI Act by applying their existing review and editorial procedures and established professional standards, as appropriate.