EU begins enforcing Article 50 transparency rules for chatbots, requiring disclosure of key system details

The European Union’s AI Act moved into day-to-day enforcement on August 2, when Article 50 transparency obligations began applying to certain AI systems used across consumer, media, education and platform settings. The rules require providers and, in some cases, deployers to tell people when they are interacting with AI, and to label or mark certain synthetic outputs, including deepfakes. The European Commission published implementation guidance in July and said the measures are meant to help users distinguish AI-generated or manipulated content from authentic material. ### Which AI tools are now covered? Article 50 of Regulation (EU) 2024/1689 covers several categories rather than one broad class of “AI content.” The law and Commission guidance say the obligations apply to AI systems intended to interact directly with natural persons, emotion-recognition or biometric-categorisation systems in certain contexts, systems that generate or manipulate image, audio or video content constituting deepfakes, and certain AI systems that generate or manipulate text published to inform the public on matters of public interest. (commission.europa.eu) The Commission said chatbots and other conversational systems generally must make clear to users that they are dealing with AI, unless that is obvious from the context. The guidance says the disclosure should be provided at the latest at the first interaction and in a clear, accessible manner. ### What exactly has to be disclosed? The Commission’s guidance says the core duty for chatbots is identity disclosure, not publication of a model card or full technical dossier to end users. (eur-lex.europa.eu) Users must be informed that they are interacting with an AI system, while downstream business users may need additional information from providers to support their own compliance. (commission.europa.eu) For synthetic media, the Commission said providers must ensure AI-generated image, audio and video outputs are marked in a machine-readable format where technically feasible, and deployers of deepfakes must disclose that the content has been artificially generated or manipulated. For AI-generated text published on matters of public interest, deployers must disclose that the text was artificially generated or manipulated unless a human has exercised editorial control and taken responsibility, or another exception applies. (digital-strategy.ec.europa.eu) ### Why are publishers, schools and creators paying attention? The Commission said the rules are aimed at risks including misinformation, impersonation, fraud and consumer deception. Because Article 50 reaches public-facing outputs, the first compliance questions have fallen on publishers, educators, platforms and creators that use generative tools in ordinary workflows, not only on frontier model developers. (digital-strategy.ec.europa.eu) WIRED reported that AI influencers and virtual creators were already reassessing business practices as the EU rules approached, with some treating disclosure as a branding choice and others worried about legal uncertainty. EdTech Innovation Hub reported that the rules took effect for chatbots, deepfakes and AI-generated content used in education and other public-facing settings on August 2. (commission.europa.eu) ### Who is supposed to comply: the model maker or the company using it? The AI Act splits duties between providers and deployers. The regulation and Commission guidance say providers are responsible for designing covered systems so required disclosures or output markings can be made, while deployers carry their own duties when they use deepfakes or publish certain AI-generated content to inform the public. That division matters for companies building on top of third-party models. (digital-strategy.ec.europa.eu) The Commission’s FAQ and outside legal analyses note that a vendor’s documentation does not automatically satisfy a deployer’s obligations if the deployer controls the user-facing interaction or publication context. ### How is Brussels trying to make the rule workable? The European Commission published Article 50 guidelines on July 20 and also backed a voluntary Code of Practice on Transparency of AI-generated Content. (eur-lex.europa.eu) The code sets out measures on marking and detection of AI-generated content and is intended as a practical compliance tool alongside the legal text and FAQs. (digital-strategy.ec.europa.eu) Enforcement now shifts to national market-surveillance authorities and other competent bodies under the AI Act framework. The Commission’s next role is likely to center on additional guidance, coordination and interpretation as providers, publishers and platforms test how the Article 50 duties work in live products after August 2. (eur-lex.europa.eu) (digital-strategy.ec.europa.eu)