EU Commission: the European Code of Transparency has been endorsed as the main tool for complying with Article 50 of the AI Act from 2 August onwards.

The European Commission and the European Committee on Artificial Intelligence have given a positive assessment of the Code of Good Practice on the Transparency of AI-Generated Content, considering it an appropriate tool to facilitate the application of the obligations set out in Article 50(2), (4) and (5) of the AI Act.

The Code, published on 10 June 2026, is a voluntary framework aimed at providers and deployers of generative AI systems. Adherence to it does not replace the obligations directly laid down by the AI Act, but offers signatories a European-level operational framework to demonstrate compliance with the rules on the marking, detection and labelling of artificial content.

The rules will come into force on 2 August 2026. From that date, providers of systems that generate or manipulate text, images, audio or video must ensure that outputs are marked in a machine-readable format and are detectable as artificially produced or modified, to the extent technically feasible. The solutions adopted must be effective, interoperable, robust and reliable.

For deployers, the Code specifically concerns the obligation to make deepfakes and AI-generated or manipulated texts recognisable when published for the purpose of informing the public on matters of general interest. For the latter, this obligation is subject to the exception provided for in the AI Act where the content has undergone human review or editorial oversight and a natural or legal person assumes responsibility for the publication.

The Code is divided into two sections. The first is intended for providers and sets out the technical measures for labelling and identifying content that has been generated or manipulated. The second concerns deployers and establishes common procedures for informing users of the artificial nature of deepfakes and certain text-based publications.

Operational measures include the use of machine-readable labels, the use of open standards, the development of interoperable detection mechanisms, and the adoption of disclosure methods that are clearly recognisable to users. The EU has also established a common system of icons that creators, publishers and other deployers can use to indicate the presence of content generated or manipulated by AI. The icons are freely available for use, but their use alone does not ensure compliance with Article 50: the deployer remains responsible for the clarity, visibility and adequacy of the information provided.

Positive assessment is particularly significant in terms of implementation. Those who adhere to the Code and correctly implement its measures will be able to use it to demonstrate compliance with the obligations of the AI Act, benefiting from greater legal certainty, uniform application across the EU and a reduction in administrative burdens. Those who opt for alternative approaches will, however, have to demonstrate their adequacy independently, which may be assessed on a case-by-case basis by national market supervisory authorities.

However, the Code does not constitute the entire regulatory framework. It will be supplemented by the Commission’s guidelines on Article 50, which are intended to clarify which entities and content fall within the scope of the obligations, how exceptions should be applied, and which labelling and marking methods are actually appropriate.

For businesses, publishers, platforms and developers, the practical guidance is therefore to begin mapping content produced or modified using AI systems as of now, to distinguish between the obligations of providers and those of deployers, to review the available labelling techniques, and to establish internal procedures for applying labels, icons and information to users.

Adherence to the Code remains voluntary, but it represents the main common approach indicated by the European institutions for translating Article 50 of the AI Act into technical and organisational measures that can be verified in practice.