The European Union has published its final Code of Practice for marking and labelling AI-generated content, adding a practical layer to the bloc’s wider AI governance framework. Released on 10 June 2026, the code is designed to help providers and deployers of generative AI comply with the AI Act’s transparency rules, especially around identifying synthetic media and informing people when they are dealing with AI systems.
Although the code is voluntary, the European Commission is presenting it as an important compliance tool a of the next legal milestone. The relevant AI Act transparency obligations will start applying on 2 August 2026, and the Commission says the code should give participating organizations more predictability, legal certainty, and a lower administrative burden across the EU.
Why the EU issued the code now
The timing is closely tied to the AI Act rollout. The Commission has long signaled that transparency obligations for generative AI would become applicable in August 2026, and the code now arrives just weeks before that date. In policy terms, the aim is to move from broad legal principles to a more operational framework that companies can actually implement.
The Commission says the code is meant to address risks of deception and manipulation in the information ecosystem. That includes concerns about deepfakes, AI-altered media, and certain AI-generated text that may influence public debate. By publishing a common framework, the EU is trying to create shared expectations for how synthetic content should be marked and disclosed.
Official Commission wording frames the initiative around a clear purpose: helping providers and deployers “comply with the AI Act’s obligations for labelling and marking of AI-generated content.” That language matters because it shows the code is not a separate policy experiment, but a supporting instrument for an existing legal regime.
What the AI Act requires from August 2026
The core legal deadline is 2 August 2026. From that date, the AI Act’s transparency rules connected to this code will apply, requiring specific disclosures in several high-risk information contexts. The rules are especially relevant for organizations that create, distribute, or use generative AI outputs in public-facing settings.
According to the Commission, the requirements include clear labelling of deepfakes and of AI-generated or AI-manipulated text on matters of public interest. The goal is to make it easier for people to understand when content has been synthetically produced or significantly altered by AI, particularly when that content could affect civic discourse, trust, or public decision-making.
The transparency framework also extends beyond media files and publications. Users must be informed when they are interacting with an AI system such as a chatbot. This reflects a broader regulatory principle: people should know when they are engaging with machine-generated communication rather than assuming they are dealing with a human source.
The two main sections of the code
The new code is organized into two major sections, each aimed at a different part of the AI value chain. Section 1 focuses on providers of generative AI systems, while Section 2 is directed at deployers that use those systems in real-world products, services, and communication contexts.
Section 1 covers the marking and detection of AI-generated or AI-manipulated audio, image, video, and text. The Commission emphasizes that these commitments should be effective, interoperable, robust, and reliable, as far as technically feasible. A key point here is the preference for machine-readable marking, which would allow automated systems and platforms to detect or preserve signals that content was generated or altered by AI.
Section 2 deals with the labelling of deepfakes and certain AI-generated or AI-manipulated text on matters of public interest by deployers. This section is more directly visible to end users, because it concerns how organizations present content labels in practice. It is also where the EU’s new visual tools for disclosure become especially relevant.
EU icons and practical labelling tools
Alongside the code, the EU has published “EU Icons for labelling AI-generated content.” The Commission describes these icons as freely available and as an integral part of Section 2. That makes them more than a design add-on: they are intended to support harmonized communication of AI disclosures across the European market.
Standard icons could help reduce fragmentation in how labels appear from one service or platform to another. If users repeatedly see the same symbols in different contexts, they may become better at recognizing synthetic or AI-manipulated content. That consistency is especially valuable in fast-moving online environments where people often make judgments in seconds.
For businesses and public organizations, the icons may also simplify implementation. Rather than inventing custom labels from scratch, deployers can rely on a set of common visual cues aligned with the EU’s policy approach. In practice, this could support both usability and compliance, especially for organizations operating in multiple Member States.
Voluntary code, but meaningful incentives
The Commission stresses that the code is voluntary. Companies are not legally forced to sign it in order to meet the AI Act, and the underlying legal obligations still come from the regulation itself. Even so, Brussels is clearly encouraging adoption by linking participation to regulatory advantages.
According to the Commission, signatories will benefit from greater predictability, more legal certainty, and reduced administrative burden across the EU. Those incentives are important in a market where organizations may otherwise face uncertainty about what good-faith compliance looks like in technical and operational terms. A common code can function as a recognized benchmark.
The Commission has also indicated that future enforcement for signatories will focus on monitoring adherence to the code. That point gives the voluntary framework extra weight. In effect, joining the code may offer companies a clearer supervisory path, while also signaling a commitment to responsible AI development and use in the EU.
How the code was developed
The final document is the result of a multi-stakeholder process launched in September 2025. The Commission says more than 187 participants took part, including representatives from industry, academia, civil society, rightsholders, EU Member States, and external observers. That broad participation was intended to balance technical feasibility with public-interest concerns.
The preparatory timeline moved in several stages. The Commission launched the work on the code on 5 November 2025, setting out early expectations and noting that the transparency obligations would apply in August 2026. The first draft followed on 17 December 2025, giving stakeholders a concrete basis for debate.
Feedback on that draft was collected until 23 January 2026, and the Commission said at the time that a second draft would follow by mid-March 2026. The final publication on 10 June 2026 therefore reflects months of consultation and revision rather than a last-minute regulatory add-on. The code is also now under adequacy assessment by the Commission and the AI Board.
Sign-ups, next steps, and market impact
The code is already open for signatures. Since 10 June 2026, any provider or deployer of a generative AI system has been able to join by submitting a signatory form through the Commission’s process. The Commission has said signatories will be publicly listed in July 2026, which may create early visibility around who is choosing to align with the framework.
The EU has also scheduled an information session for 22 June 2026, and the signing page notes that questions can be sent to the AI Office email address provided there. These steps suggest the Commission is not just publishing the code and stepping back; it is actively trying to build uptake before the August compliance date.
For the market, the practical effect may be significant. Providers will need to think carefully about machine-readable marking and detection methods, while deployers will need workflows for applying visible labels to deepfakes and relevant AI-generated text. In both cases, the new EU code for AI content labeling could become an important reference point for procurement, platform policy, and public-sector communication.
The publication of the code marks another step in the EU’s effort to turn AI regulation into day-to-day operational practice. Rather than relying only on legal obligations written at a high level, the Commission is offering a structured method for companies and organizations to show how they label, mark, and disclose AI-generated content in real settings.
Whether the voluntary code becomes a widely adopted standard will depend on industry uptake in the weeks a. But with the AI Act’s transparency rules taking effect on 2 August 2026, the direction of travel is already clear: in Europe, AI-generated content will increasingly need to be identifiable, explainable, and visibly labelled for the people who encounter it.
