The European Union has published its long-awaited AI content labelling playbook, and it's changing how millions of people will interact with artificial intelligence from August.
The European Commission released the voluntary Code of Practice on 10 June. It provides practical guidance for developers and deployers of generative AI systems on how to meet transparency obligations that become enforceable across the bloc from 2 August 2026.
While participating in the Code is optional, the legal requirements underneath it aren't. Under Article 50 of the EU AI Act, organisations must clearly disclose when users are interacting with AI-powered systems — think virtual assistants, customer-service chatbots, or automated support agents.
Certain categories of AI-generated content must also carry visible labels indicating their artificial origin. The rules pay particular attention to deepfakes and AI-generated content related to matters of public interest — synthetic images, audio recordings, videos, and text that could influence public debate or shape perceptions of social, political, or economic issues.
The framework splits responsibilities. Developers of generative AI systems are encouraged to embed machine-readable markers into AI-generated outputs. These technical identifiers make synthetic content easier to detect and trace as it moves across digital platforms and communication channels.
Organisations deploying AI technologies bear responsibility for ensuring users receive visible and understandable disclosures when required by law. The Commission recognises that AI-generated content often passes through multiple stages before reaching the public, so transparency measures must be implemented throughout the AI supply chain rather than relying on a single point of control.
To support consistency across member states, the Commission is promoting open technical standards alongside a common European visual icon that can be used to identify AI-generated material. This approach aims to provide a recognisable signal for users while reducing compliance complexity for businesses operating across multiple markets.
The guidance also acknowledges the role of human oversight. For AI-generated text concerning matters of public interest, labelling requirements generally apply where content is published without meaningful human review or editorial intervention.
Although the Code is now open for signatures, businesses face a challenging timeline. Companies serving European customers have less than two months to evaluate their AI systems, identify where transparency obligations apply, and implement appropriate labelling mechanisms before enforcement begins.
For many organisations, compliance may involve reviewing existing workflows, updating customer-facing interfaces, integrating technical watermarking solutions, and establishing governance processes to monitor AI-generated content.
The challenge is compounded by the fact that several important implementation details remain under development. The European Commission is expected to publish additional guidance clarifying aspects of the legislation and addressing areas not fully covered by the Code of Practice. This leaves many businesses balancing regulatory preparation with a degree of uncertainty regarding how specific provisions will ultimately be interpreted and enforced.
Technology firms, media organisations, public institutions, and digital platforms are among the sectors expected to be most directly affected by the new requirements.
The EU AI Act has been in the works for years, and this labelling playbook is one of its most concrete outputs. European policymakers argue that transparency is essential for maintaining public trust as AI technologies become increasingly integrated into everyday life. According to the Commission, effective labelling can help reduce the risks of misinformation, manipulation, and deception while allowing businesses to continue innovating responsibly.