The EU's Code of Practice on AI-Generated Content Commits Signatories Other Than Media Service Providers to Policies for Human Review or Editorial Control Before Publication

The EU's AI Act sets transparency obligations for generative AI in its Article 50, and the European Commission's page on the code that supports them says those obligations are "applicable from 2 August 2026." The Code of Practice on Transparency of AI-Generated Content was published in its final form on June 10, 2026, after a process the page says was facilitated by the AI Office. Parts of its second section, on deployers of AI systems, follow, not the whole code.

Voluntary Code, Legal Obligation

"Even though adherence to the code is voluntary, the transparency requirements under article 50 of the AI Act are legal obligations," the Commission's page says. The Commission and the AI Board, it says, have confirmed that the code is an adequate voluntary tool to demonstrate compliance, while providers and deployers that decide to comply through other means will have to demonstrate that those measures are adequate.

One of the stated objectives of the code's section for deployers is "to serve as a guiding document for demonstrating compliance with the obligations of deployers of generative AI systems provided for in Article 50(4) and (5) AI Act, while recognising that adherence to the Code does not constitute conclusive evidence of compliance with these obligations under the AI Act."

Voluntary, and not conclusive.

Published Text and the Exception

The law's text, as the code reprints it, says a deployer of an AI system that generates or manipulates text "which is published with the purpose of informing the public on matters of public interest shall disclose that the text has been artificially generated or manipulated." The obligation does not apply, among other cases, "where the AI-generated content has undergone a process of human review or editorial control and where a natural or legal person holds editorial responsibility for the publication of the content."

The code's section for deployers applies to signatories only in so far as they are deployers of AI systems that generate or manipulate deep fakes, or such text. In that section, "published text" implies AI-generated or manipulated text published to inform the public on matters of public interest, "without human review or editorial control and where no natural or legal person holds editorial responsibility for the publication of the content."

Human Review and Editorial Control

The code's Commitment 4 covers that review. Signatories who are media service providers "within the meaning of Art. 2(2) of Regulation (EU) 2024/1083," already subject to editorial standards and frameworks, may rely on the exception by applying their existing review and editorial procedures and established professional standards. All other signatories "commit to establish, adapt, or maintain appropriate policies for human review or editorial control prior to publication and that a natural or legal person holds editorial responsibility for the publication."

Those policies may rely on existing processes and will be proportionate to the deployer's size and resources. They will include at least the identification of the natural or legal person with editorial responsibility, with "name, role and contact details," and an overview of the concrete organizational measures and human resources allocated to ensure that adequate human review or editorial control is performed, and editorial responsibility assumed, before publication. That "does not entail having to document individual instances of human review or editorial control over individual text publications." Where not already publicly available, signatories commit to publish the contact details of those with editorial responsibility "to ensure accountability," and they may record additional information on the nature of the review.

The code asks for an identified natural or legal person and a described process, and says each review need not be documented. For a reader of a single text, that leaves open how to tell whether that text was in fact reviewed.