EU’s New Code: Meeting the August Deadline

The European Union has recently unveiled its AI content labelling playbook, a voluntary Code of Practice designed to assist companies in complying with transparency regulations that will come into effect across the bloc on August 2. This Code, released by the European Commission on June 10, outlines practical steps for businesses engaged in building and utilizing generative AI to appropriately mark and label the content generated by their systems.

Although the Code itself is optional, the obligations it refers to are mandatory and fall under Article 50 of the EU AI Act. Starting from August 2, 2026, these obligations will apply regardless of whether a company signs the Commission’s guidance. Signing the Code simply provides a recognized method for demonstrating compliance.

The key requirements of the AI content labelling rules, effective from August onwards, include the clear identification of deepfakes and AI-generated or manipulated text related to public interest. Additionally, users engaging with interactive AI systems, such as customer service bots, must be informed that they are interacting with a machine.

The rationale behind these rules is to empower users in identifying AI-generated or altered content and to reduce opportunities for deception. Henna Virkkunen, the Commission’s executive vice-president for tech sovereignty, security, and democracy, emphasized the importance of transparency in content creation using AI, especially in shaping public discourse.

The Code provides a practical framework for AI providers and deployers to implement labelling measures before the regulations take effect in August. It divides the responsibility between companies developing generative models, who are tasked with marking their output in a machine-readable format, and those deploying these models in real products, responsible for visible labelling.

To ensure consistency, the Code advocates for the use of open technical standards and a common EU icon as visual cues for users. Providers and deployers are encouraged to sign the Code, which is currently open for signatures and awaiting evaluation by the Commission and the AI Board. Separate guidelines from the Commission are expected to clarify the law and address any gaps left by the Code.

With less than two months before the regulations come into force, companies serving European users must swiftly determine their labelling requirements and decide whether to sign the Code. Detailed guidelines from the Commission are yet to be released, adding to the urgency of compliance.

In conclusion, the Code represents a significant step towards enhancing transparency in AI content creation and marking a milestone in addressing AI content labelling under the EU AI Act. It underscores the importance of user awareness and industry accountability in the era of AI-driven content generation.