European Commission publishes second draft code for labelling AI-generated content
The European Commission has released a revised draft of a voluntary code of practice designed to help organisations comply with transparency requirements for AI-generated content under the EU AI Act.
The European Commission has published the second draft of a Code of Practice on marking and labelling AI-generated content, part of preparations for implementing transparency requirements in the EU Artificial Intelligence Act.
The voluntary code is intended to help companies and organisations understand how to meet the obligations of Article 50 of the AI Act, which requires certain types of AI-generated content to be clearly identified. These rules aim to ensure that users can recognise when text, images, audio, or video have been created or manipulated by AI systems.
The updated draft was prepared by independent experts and incorporates feedback from hundreds of stakeholders, including representatives from industry, academia, civil society organisations, EU member states, and members of the European Parliament. Input was gathered through surveys, workshops, and consultations held in early 2026.
The document is divided into two main sections. The first focuses on providers of generative AI systems, outlining technical methods for marking AI-generated content. The revised draft proposes a two-layer approach combining metadata and watermarking, with optional techniques such as fingerprinting and logging to support detection and verification.
The second section addresses organisations that deploy AI systems, such as platforms publishing or distributing content. It focuses on the labelling of deepfakes and AI-generated material related to matters of public interest. The draft includes guidance on the design and placement of icons, labels, or disclaimers that inform users when content has been created by AI.
The revised version simplifies earlier proposals, introduces greater flexibility for organisations implementing the rules, and encourages the use of open technical standards. The annex also includes examples of a potential EU icon that could be used to mark AI-generated content across platforms.
The Commission is collecting feedback on the draft until 30 March 2026, with the final version expected by June 2026. The transparency requirements for AI-generated content under the AI Act are scheduled to apply from 2 August 2026.
