The EU imposes strict rules for AI content before August 2026

The European Union recently unveiled its guide for labeling content generated by artificial intelligence, a voluntary Code of Practice aimed at preparing businesses for new transparency requirements. These rules will come into effect across the EU starting August 2, 2026. The European Commission published this final Code on June 10, providing practical guidelines for companies developing and using generative AI to mark and label the outputs of their systems.

Although the Code is optional, the obligations it mentions are not. They stem from Article 50 of the EU AI Act. From August 2, 2026, these obligations will apply to all companies, whether or not they have signed the Commission's recommendations. Signing the Code simply offers a recognized way to demonstrate compliance.

Labeling Requirements for AI Content

Starting in August, two elements must be clearly indicated: deepfakes and texts generated or manipulated by AI on topics of public interest must carry a label. Additionally, anyone interacting with an interactive AI system, such as a customer service bot, must be informed that they are dealing with a machine.

The European Commission presents these measures as a way to help users identify material created or modified by AI, thereby reducing the risks of deception. Henna Virkkunen, Executive Vice-President of the Commission for Technological Sovereignty, Security, and Democracy, emphasized that Europeans have the right to know whether what they see, hear, or read has been created or modified by AI, especially when this content can influence public debate.

The Code offers a practical means of labeling that AI providers and deployers can follow before the rules come into effect in August. It divides responsibilities between the two parties in the AI supply chain. Companies that build generative models are encouraged to mark their output in a machine-readable format, allowing for its subsequent detection in the process.

Companies that deploy these models, those integrating AI into real products, are responsible for visible labeling. This applies to AI-generated text of public interest disseminated without human review or editorial oversight. To facilitate enforcement, the Code relies on open technical standards and a common EU icon, providing users with a consistent visual cue and preventing companies from creating their own solutions.

An Evolving Process

The Code is now open for signatures, and the Commission encourages all providers and deployers to commit. It is up to the Commission and the AI Council to assess its adequacy, while separate guidelines from the Commission still need to clarify certain aspects of the law not covered by the Code. Developed by six independent experts with contributions from over 180 stakeholders, this Code is the first instrument to address the labeling of AI content within the framework of the law.

With a tight timeline, companies serving European users have less than two months to determine what they need to label and how, and to decide whether they wish to sign the Code. Many complex details still depend on guidelines that the Commission has yet to publish.