EU AI labeling law takes effect Sunday: chatbots, deepfakes, and AI-altered images now require disclosure

EU AI Act Article 50 transparency rules took effect Sunday, requiring companies that distribute AI-made content in the EU to label it. The obligations land in phases: chatbot disclosure first, with deepfakes and AI-generated or altered images following.

Providers must apply machine-readable marking to AI-generated or manipulated audio, image, video, and text. Deployers must label deepfakes and flag AI-generated text publications. Chatbots must tell users they are speaking with an AI. The Commission has spelled out the scope in its Code of Practice on AI-generated content, and the statutory text lives in Article 50 of the AI Act.

Personal, non-professional use is exempt, according to a Greenberg Traurig analysis of the Commission Code of Practice. Companies that miss the rule face fines under the Act's Article 99 penalty regime, which analysts describe as steep. A Greenberg Traurig explainer treats the regime as a disclosure floor rather than a content ban.

An EU official, quoted in coverage of the rollout, said generative AI "enables disinformation to be created at unprecedented scale, tailored to specific audiences, and disseminated with remarkable speed," and that the rules aim to "preserve citizens' ability to trust what they see, hear, and read."

Open questions: how aggressively national regulators will police chatbot disclosure in the first weeks, and whether the Commission will name compliant platforms before deepfake labeling fully phases in.