.jpg?w=1024)
AI-generated content is getting a rulebook in Europe… and companies now have a clearer map for how to label it.
The European Commission has published the final Code of Practice on marking and labelling AI-generated content, setting out how companies should prepare for upcoming legal transparency rules under the EU AI Act. The Code is voluntary, but it is designed to help providers and deployers of generative AI systems comply with binding obligations that will take effect in August 2026.
From that date, the EU AI Act will require clear labelling of deepfakes and AI-generated or manipulated text published on matters of public interest. It will also require that users be informed when interacting with AI systems, such as chatbots.
According to the Commission, these measures are intended to reduce risks of deception and manipulation while improving transparency in digital spaces.
What the new code covers
The Code of Practice was drafted by six independent experts with input from more than 180 stakeholders, including AI companies, SMEs, academics, public sector groups, and civil society organisations.
It is split into two main sections: One focuses on providers of generative AI systems, and the other on deployers, meaning those who use AI tools to publish or distribute content. For providers, the Code outlines how AI-generated audio, images, video, and text should be marked in a machine-readable way so that they can be detected as synthetic or manipulated.
For deployers, it focuses on how deepfakes and AI-generated text published in the public interest must be clearly labelled when no human editorial review is involved.
There are limited exemptions, including cases where content is clearly artistic, satirical, or fictional, or where AI is used under legal authority, such as in criminal investigations.
“Europeans have a right to know whether what they see, hear or read has been made or altered by AI, especially when such content can shape public debate. Transparency is how we protect trust. This Code of Practice gives AI providers and deployers a clear and practical way to label AI-generated content and deepfakes in line with the AI Act’s rules that will apply from 2 August,” said Henna Virkkunen, Executive Vice-President for Tech Sovereignty, Security and Democracy.
EU also introduces official AI labels
Alongside the Code, the EU has developed a set of official icons for labelling AI-generated content.
These include variations for fully AI-generated content, partially modified content, and a basic AI marker that can be adapted with additional text or interface elements. The icons are designed to be clearly visible and to appear at first exposure to the content, whether in images, video, or text.
For video content, labels must appear at the start and at intervals. For text, they must be placed near headlines or at the beginning of a publication. Audio content must include a spoken disclaimer in plain language, with repeated reminders for longer formats.
Voluntary Code, mandatory law
While the Code of Practice itself is voluntary, the underlying AI Act requirements are not.
Companies that sign the Code and follow its measures may more easily demonstrate compliance with EU law. Those who do not sign will still be required to meet the same legal standards but may have to prove compliance individually.
The Code is now open for signatures from AI providers and deployers. If approved by the Commission and the AI Board, signatories can use it as proof of compliance with EU rules. The Commission also plans to issue further guidelines to clarify how the transparency rules should be applied in practice.
The next test will be whether major AI providers and content platforms sign on before the 2026 rules take effect. Even without signatures, the AI Act’s transparency requirements are coming, which means companies distributing AI-generated public-interest content will need a defensible labelling strategy before the deadline.
Also read: The EU is expanding its tech sovereignty agenda with new cloud, AI, and chip measures aimed at reducing dependence on foreign tech providers.