EU AI Act rules bring new labels and disclosures for synthetic content across Europe. Entities offering systems that interact directly with individuals must inform users when they are engaging with AI. Such notices are not necessary if the artificial origin of the interaction is obvious to a reasonably informed person. Providers of systems that generate synthetic text, audio, images, or video are also required to include machine-readable indicators. These marks should enable detection of AI-created or manipulated content whenever technically feasible. Basic editing tools that do not significantly alter the input or its meaning are exempt from this requirement. The law distinguishes between these technical marking responsibilities and the visible disclosures that deployers must provide to their audiences. Deployers must identify deepfake images, audio, and video when the material could falsely seem authentic. They are also tasked with labeling AI-generated texts published for the public, especially on issues of public interest. However, this text labeling obligation does not apply after a human review or editorial oversight, provided the responsible individual or organization accepts editorial accountability. All required disclosures must be presented clearly, in an understandable and accessible manner. Distinctive machine markings are separate from public-facing labels.
