The European Commission Specifies Labelling Requirements for AI-Generated Content and Deepfakes

With the transparency obligations under Article 50 AI Act, the European legislator has, for the first time, created union-wide disclosure and labelling obligations for certain AI-generated content. This provision is gaining increasing practical relevance, particularly in the field of generative AI. This is because modern AI systems now make it possible, without significant technical effort, to create deceptively realistic image, audio, video, and text content, the artificial origin of which is often not readily apparent to users.

While the provision has so far primarily set out abstract requirements, the European Commission is now specifying its practical implementation. Together with a new “Code of Practice on Transparency of AI-Generated Content,” the Commission published, on 10 June 2026, official EU icons for the first time, which can be used to label AI-generated or AI-manipulated content. The Commission makes the icons, together with further information on the practical implementation of the transparency obligations, publicly available on its official website.

I. Voluntary Code of Practice Instead of Binding Requirements

The Code of Practice, however, is expressly not binding law. The Code is designed to be voluntary and is intended to provide providers and deployers of generative AI systems with practical guidance on implementing the transparency obligations under Article 50 AI Act. The actual statutory labelling obligations continue to derive directly from the AI Act itself and apply from 2 August 2026.

Nevertheless, the new guidelines are likely to be of considerable practical significance. This is because, for the first time, the Commission is specifying here how it envisages the technical and visual implementation of the transparency obligations. This concerns, in particular:

• the design of labels,
• their visibility and placement,
• their permanent recognisability when shared or downloaded,
• as well as requirements regarding accessibility and machine-readable notices.

From 2 August 2026, Article 50 AI Act requires, in particular, clear labelling of:

• deepfakes,
• AI-generated or AI-manipulated image, audio, and video content,
• as well as certain AI-generated texts on matters of public interest.

In addition, users must be informed when they are communicating with interactive AI systems such as chatbots or virtual assistants.

II. The New EU Icons for Labelling AI-Generated Content

The newly published EU icons are intended to help make such content “clearly perceivable and distinguishable” to users as artificially generated or manipulated. In particular, the following are envisaged:

• a basic icon for general AI involvement,
• a label for fully AI-generated content,
• as well as a symbol for partially AI-manipulated content.

III. Voluntary Signing of the Code of Practice

Alongside the publication of the new transparency guidelines, the European Commission’s AI Office has explicitly invited providers and deployers of generative AI systems to voluntarily sign the “Code of Practice on Transparency of AI-Generated Content.”

By signing, companies express that they intend to take the standards contained in the Code into account going forward as practical guidance for implementing the transparency obligations under Article 50 AI Act. According to the Commission, this is intended not only to contribute to greater legal and planning certainty within the European Union, but also to noticeably reduce the subsequent compliance burden. The Commission’s announced supervisory practice deserves particular attention in this regard.

According to the Commission, signing the Code is intended to enable greater predictability, more legal certainty within the European Union, and a reduced administrative compliance burden. Future enforcement measures are intended to be primarily guided, with respect to signatories, by compliance with the Code. Companies that sign the Code of Practice could thereby benefit from a union-wide simplified and more harmonised compliance practice, regardless of the Member State in which they operate or which market surveillance authority they fall under.

According to the Commission, the signatories of the Code are to be publicly announced as early as July 2026, and thus even before the official date on which the transparency obligations under Article 50 AI Act take effect on 2 August 2026. For signing, the European Commission provides an official “Signatory Form.” Providers and deployers of generative AI systems can voluntarily support the Code by submitting the signed form to the AI Office and thereby join the transparency standard.

IV. Outlook

Although the European Union’s new Code of Practice remains voluntary, it is likely to significantly shape the practical implementation of the transparency obligations under Article 50 AI Act going forward. Signing can offer advantages in terms of legal and planning certainty, but at the same time may also raise expectations regarding a company’s own compliance. How strongly the Code will influence regulatory practice and market standards in the future remains to be seen.