Rules requiring artificial intelligence content to be labelled became enforceable across the European Union on 2 August 2026, with the European Commission’s AI Office and national authorities taking on enforcement.
Under the transparency provisions of the AI Act, deepfakes — images, video or audio generated or edited using AI — must be labelled. AI-generated or altered content must also carry machine-readable marks so that it can be detected automatically. Chatbots and other interactive systems must tell users they are not dealing with a human.
The obligation applies regardless of intent. Content that appears to depict a real person must be labelled even where there was no attempt to deceive, and, according to the Commission’s guidance, even where no real individual is depicted.
Around 190 companies and organisations had signed the accompanying Code of Practice on transparency of AI-generated content by the end of July. Signatories can rely on its measures to demonstrate compliance with the labelling and detection requirements.
The reach of the rules over technology-facilitated gender-based violence — abuse committed, assisted or amplified by digital tools and directed at people because of their gender — is untested. The provisions address disclosure rather than creation or distribution, and a labelled non-consensual sexual image remains a non-consensual sexual image.
Specialists have questioned whether the rules can be enforced as drafted. Analysis published by Euronews on 28 July reported warnings that technical gaps and the absence of common standards for marking content could undermine enforcement in practice.
The rules took effect two days after a federal judge in Minnesota allowed the first US state ban on nudification tools to come into force.