Commission Opinion of 8 July 2026 on the assessment of the Code of Practice on Transparency of AI-generated Content — C(2026) 4839 final

2. BACKGROUND

(8) Codes of practice play an important role in the implementation of the AI Act. While they remain voluntary, the Union legislature considers such codes as an effective means to facilitate compliance by providers and deployers of AI systems and AI models with their obligations laid down in the AI Act. The Code sets out commitments and measures to which providers of generative AI systems, including general-purpose AI systems, and deployers of AI systems generating deepfakes and certain text may adhere to demonstrate their compliance with the obligations laid down in Articles 50(2) and 50(4) AI Act respectively and the horizontal disclosure requirements in Article 50(5) AI Act.

(9) The Code was prepared following an iterative drafting process.(1) On 5 November 2025, that process was commenced with the ‘Kick-off Event for the Code Plenary’ involving more than 180 participants representing eligible respondents to the call to participate in the drawing-up of the Code, launched by the AI Office in early September 2025. A variety of interested stakeholders were involved in the drafting process, including providers of generative AI systems, developers of marking and detection techniques, associations of deployers of generative AI systems, civil society organisations, academic experts, standardisation organisations and specialised organisation with expertise in transparency, and providers of very large online platforms and of very large online search engines designated in accordance with Article 33(4) of Regulation (EU) 2022/2065. A list of those stakeholders and minutes of all meetings and workshops organised for the preparation of the Code are published on the Commission website.

( 1) See more details on Code of Practice on marking and labelling of AI-generated content | Shaping Europe’s digital future

EN 2 EN

The close involvement of Member States representatives as observers, as well as through the Board and its dedicated subgroup on transparency, was ensured throughout the process. The AI Office also invited other public bodies and agencies from all over the world working on transparency of AI-generated content to participate in the Code process as observers.

(10) The Code’s drafting process involved two working groups chaired by independent, academic chairs and vice-chairs with renowned expertise in the respective areas. Working Group 1 led Section 1 of the Code on Rules for the marking and detection of artificially generated and manipulated content applicable to providers of generative AI systems (Articles 50(2) and (5) AI Act). Working Group 2 led Section 2 of the Code on Rules for labelling deep fakes and artificially generated and manipulated published text applicable to deployers of AI systems (Articles 50(4) and (5) AI Act).

(11) The Code sets out eight commitments, divided into two categories. Section 1 of the Code operationalises the obligations in Articles 50(2) and (5) AI Act with four commitments on marking and ensuring detectability of AI-generated or manipulated content applicable to providers of AI systems generating or manipulating synthetic audio, image, text or video content. Section 2 of the Code operationalises the obligations in Articles 50(4) and (5) AI Act with four commitments on labelling deepfakes and AI-generated or manipulated text published with the purpose of informing the public on matters of public interest (referred to as ‘published text’) applicable to deployers of AI systems.

(12) Under Section 1, Commitment 1 sets out the marking of AI-generated or manipulated content with a set of measures. Commitment 2 sets out measures for the detection of markings of AI-generated or manipulated content. Commitment 3 describes the measures to meet the requirements for marking and detection solutions. Commitment 4 addresses testing, verification and compliance.

(13) Under Section 2, Commitment 1 sets out three measures to facilitate the disclosure of deep fakes and published text. Commitment 2 foresees three measures concerning internal processes for i) compliance, ii) awareness raising and literacy, and iii) review, feedback and cooperation with competent authorities. Commitment 3 sets out measures to operationalise the special regime for disclosure of deep fakes that are part of artistic, creative and similar works or programmes. Commitment 4 provides measures to ensure human review, editorial control and editorial responsibility for text published with the purpose to inform the public on matters of public interest applicable to deployers who wish to rely on the exception from the labelling obligation in Article 50(4) AI Act.

3. ASSESSMENT

3.1. Assessment criteria

(14) The Commission must assess the adequacy of the Code pursuant to Article 50(7) AI Act. The purpose of the Code is to facilitate the effective implementation of the obligations regarding the detection and labelling of artificially generated or manipulated content laid down in Articles 50(2), (4) and (5) AI Act. The assessment is therefore whether the commitments set out in the Code ensure the “effective implementation” of those obligations.

EN 3 EN

(15) Recital 135 AI Act specifies, in this respect, that the Code should support practical arrangements for making, as appropriate, the detection mechanisms accessible and facilitating cooperation with other actors along the value chain, disseminating content, or checking its authenticity and provenance to enable the public to effectively distinguish AI-generated content.

(16) The facilitation of an effective implementation of Articles 50(2), (4) and (5) AI Act is assessed in the following sections 3.2 to 3.6 of this Opinion.

3.2. Contribution to the effective implementation of Article 50(2) and (5) of the AI Act

(17) Section 1 of the Code aims to contribute to the effective implementation of the obligations imposed on providers of generative AI systems under Articles 50(2) and (5) AI Act.