EU publishes Code of Practice for labelling AI-generated content

What this is about

On 10 June 2026, the European Commission published the final Code of Practice on marking and labelling of AI-generated content. The code is voluntary and sets out practical steps to help providers and professional deployers of generative AI systems meet the transparency obligations of the AI Act. Those obligations, set out in Article 50 of the regulation, apply from 2 August 2026 — so the publication lands roughly eight weeks before the deadline.

The final text follows several draft rounds since the beginning of 2026, during which industry, civil society and academia submitted feedback. Alongside the code, the Commission released official EU icons for labelling AI content and a signature process through which companies can publicly commit to the code.

What the Code of Practice actually does

The code translates the AI Act's abstract requirements into concrete working steps for two groups. Providers of generative AI systems are expected to mark their outputs in a machine-readable format, for instance through metadata or watermarking techniques, so that other systems can detect that a piece of content was generated or manipulated by AI. Deployers — organisations using generative AI professionally — are expected to clearly and visibly label two cases: deepfakes, and AI-generated or AI-manipulated text published on matters of public interest.

In addition, people must be informed when they are interacting with an interactive AI system such as a chatbot. The code provides wording and design guidance for this, and the new EU icons offer a consistent visual language that users should be able to recognise across platforms.

Why it matters

From 2 August 2026, the AI Act's transparency rules are binding — regardless of whether a company signs the code. But organisations that implement the code follow a documented method aligned with the Commission, which reduces interpretation risk at a time when there is no case law on Article 50 yet.

For the wider market, the move is also a signal against deception: deepfakes and automatically generated text on political or societal topics are increasingly hard to distinguish from human content. The Commission is betting that consistent marking and labelling will reduce the risk of manipulation without banning generative AI as a tool.

In plain language

Think of the code as the ingredients list on food packaging. Nobody forbids a bakery from using ready-made dough — but the label has to say what's inside so customers can decide for themselves. In the same way, digital content should reveal whether an AI helped "bake" it: invisibly, in machine-readable form inside the content itself, and visibly, as a label for humans.

A practical example

A mid-sized marketing agency in Stuttgart produces around 200 pieces of content per month: product copy, social media clips and, occasionally, videos with virtual presenters. From 2 August 2026, it must label the virtual-presenter videos as AI-generated, since they qualify as deepfake-like depictions. If the agency adopts the code, it adds two steps to its production pipeline: the video tool embeds machine-readable provenance data, and the publishing team places the EU icon visibly in the clip. Effort after setup: a few minutes per item — far cheaper than a later dispute with a supervisory authority.

Scope and limits

Three sober caveats. First: the code is voluntary and is not legal advice; signing alone does not automatically protect against sanctions if implementation stays patchy. Second: technical marking such as watermarking is not tamper-proof — actors intent on deception can strip markings or use systems outside the EU that never mark content in the first place. Third: the code governs labelling, not detection — it offers no remedy for unmarked content from third countries, and how national supervisory authorities will enforce the rules from August 2026 remains open.

SEO & GEO keywords

EU AI Act, Code of Practice, AI content labelling, AI transparency, Article 50 AI Act, deepfake labelling, European Commission, AI watermarking, EU AI icons, generative AI compliance, AI Act 2 August 2026, AI regulation Europe