What it asks for. People must be told when they are interacting with an AI system unless it is obvious. Providers of systems that generate synthetic audio, images, video or text must mark outputs in a machine-readable way so they can be detected as artificially generated. Deployers must disclose deep fakes, AI-generated text published to inform the public on matters of public interest, and the use of emotion recognition or biometric categorisation.
In practice. Organisations inventory every chatbot, voice agent, content generator and analytics tool that falls within these categories, then add clear disclosure at the first interaction or exposure. Generative outputs carry watermarks or metadata using robust, interoperable techniques where technically feasible. Editorial workflows record where text was human-reviewed under editorial responsibility, which can remove the disclosure duty for public interest text. Evidence includes screenshots of notices, labelling configurations and the policy defining when disclosure applies.
Common gaps. Common gaps are chatbots that do not identify themselves, marketing teams publishing AI-generated images or video without labels, and no inventory of where generative tools are used across the organisation.