EU Unleashes AI Content Labeling Playbook: AI Act Deadline Looms!

The European Union has officially published its AI content labelling playbook, a voluntary Code of Practice designed to assist companies in adhering to new transparency rules. These rules are set to become legally binding across the bloc starting August 2, 2026. The European Commission released the final Code on June 10, outlining practical steps for businesses involved in building and utilizing generative AI to appropriately mark and label the content produced by their systems. While the Code itself is optional for companies to sign, the obligations it references are not. These critical requirements are established under Article 50 of the EU AI Act and will apply universally from the aforementioned date, irrespective of whether a company formally adopts the Commission’s guidance. Signing the Code merely provides businesses with a recognized method to demonstrate their compliance with the forthcoming regulations.

From August 2, 2026, two key types of AI-generated content must be clearly flagged. Firstly, deepfakes and any AI-generated or AI-manipulated text published on matters of public interest are mandated to carry a specific label. Secondly, users engaging with interactive AI systems, such as customer-service bots, must be explicitly informed that they are interacting with a machine. The Commission frames these measures as essential tools to empower users to identify material created or altered by AI, thereby significantly narrowing the scope for deception. Henna Virkkunen, the Commission’s executive vice-president for tech sovereignty, security, and democracy, emphasized the importance of these rules, stating, “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.” She presented the Code as a practical pathway for AI providers and deployers to implement labelling practices proactively before the mandatory rules take full effect.

The Code strategically divides responsibilities along the AI supply chain to ensure effective implementation. Companies that are responsible for building generative AI models are instructed to mark their output in a machine-readable format. This ensures that the AI-generated content can be detected and tracked further along the digital pipeline. Conversely, companies that deploy these models – those integrating AI into real-world products – are tasked with handling the visible labelling. This visible labelling is particularly crucial for public-interest AI text that has been disseminated without any human review or editorial oversight. To maintain workability and consistency, the Code advocates for the use of open technical standards and a common EU icon. This standardized visual cue aims to provide users with a consistent signal across different platforms and exempt businesses from the burden of creating their own unique labelling indicators.

It is important to note that the current Code is not the definitive final word on AI content labelling. It is presently open for signatures, and the Commission is actively encouraging all relevant providers and deployers to formally endorse it. The Code still requires adequate judgment from both the Commission and the AI Board. Furthermore, separate Commission guidelines are anticipated to be published. These forthcoming guidelines will aim to clarify the intricacies of the law and address any aspects that the current Code might not fully cover. Developed through the collaborative efforts of six independent experts and inputs from over 180 stakeholders, this Code represents the first instrument specifically designed to address AI content labelling under the new EU AI Act. The timeline for compliance is notably tight, with companies serving European users having less than two months to determine their specific labelling requirements, devise implementation strategies, and decide whether to sign the Code, all while awaiting further crucial details from the Commission’s upcoming guidelines.