On Wednesday 10 June the European Commission published the final version of the voluntary Code of Practice designed to help companies identify and label content generated by artificial intelligence (AI).
From 2 August 2026, providers and deployers of AI systems will be required, under the transparency obligations laid down in the AI Act, clearly to indicate content generated by AI as well as ‘deepfakes’ relating to matters of public interest that have not undergone any human review or editorial control, in order to reduce the risks of deception and manipulation.
European Commission Executive Vice-President for Tech Sovereignty, Henna Virkkunen, called on all AI providers and developers “to show leadership in developing and deploying responsible and trustworthy artificial intelligence” by signing up to this Code. “Europeans have the right to know whether what they see, hear or read has been created or altered by artificial intelligence, particularly where that content is liable to influence public debate”, she added.
For providers of generative AI systems,the Code specifies how audio content, images, videos or texts generated or manipulated by AI can be marked in a machine-readable way and detected as having been artificially created or altered, in order to comply with the obligations laid down in the AI Act.
When it comes to those deploying AI systems, the Code explains how clearly to identify ‘deepfakes’ as well as texts generated or manipulated by AI and published for the purpose of informing the public on matters of general interest, where no human review or editorial control has been exercised over their content.
To provide further assistance to deployers, the Code suggests a series of icons that they may use freely to label AI-generated content, including one icon indicating that a ‘deepfake’ was created using AI, another intended to identify content generated entirely by AI, and a third making it possible to indicate that content has been partially altered by artificial intelligence.
The Code, which is currently undergoing an adequacy assessment by European Commission and the AI Board, was drawn up by six independent experts.
Once approved by the Commission and the AI Board as complying with the requirements of the Regulation, providers and deployers signing up to it will be able to demonstrate their compliance with the relevant obligations under the AI Act. In addition, the Commission will complement this Code with guidelines intended to clarify the scope of the legal obligations and address aspects not covered by it.
Providers and deployers choosing to meet the transparency requirements by other means will have to be able to demonstrate that those measures offer an equivalent level of protection. To that end, the various national market surveillance authorities will have to carry out an individual assessment of those arrangements.
See the Code: https://aeur.eu/f/mag (Original version in French by Ana Pisonero Hernández)