EU publishes the final Code for labelling AI-generated content
New measures support compliance with upcoming EU AI transparency requirements.
The European Commission has published the final Code of Practice on marking and labelling AI-generated content, offering practical guidance for providers and deployers preparing to comply with transparency obligations under the EU AI Act.
The code is voluntary, but the underlying transparency obligations in Article 50 of the AI Act will apply from 2 August 2026. The Commission said the code is intended to help organisations implement those obligations in a consistent, practical and proportionate way.
The framework covers two main areas. Providers of generative AI systems are guided on marking and detecting AI-generated or manipulated audio, image, video and text content, including through machine-readable solutions where technically feasible. Deployers are guided on labelling deepfakes and AI-generated or manipulated text published to inform the public on matters of public interest.
Under the AI Act, users must also be informed when they are interacting with interactive AI systems, such as chatbots. The transparency requirements are intended to help people recognise when content has been generated or altered by AI and to reduce the risk of deception and manipulation.
The Commission has also published a set of the EU icons that deployers may use to label certain AI-generated content. The code does not replace the AI Act or future Commission guidelines on Article 50, which are expected before the transparency obligations begin to apply.
The Commission and the AI Board will now assess the code’s adequacy. If assessed positively, providers and deployers who sign the code may use its measures to help demonstrate compliance with the AI Act’s transparency rules.
Why does it matter?
The code is an important step in turning the AI Act’s transparency provisions into operational practice. Labelling and machine-readable marking rules could shape how platforms, AI providers, media organisations and other deployers handle synthetic text, images, audio and video. The measures are especially relevant for public-interest information, where undisclosed AI-generated or manipulated content can affect trust, elections, journalism and public debate.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!