As of Aug. 2, any chatbot, AI voice assistant or AI companion operating in the European Union must tell users, in plain terms, that they are talking to a machine. The requirement comes from Article 50 of the EU's Artificial Intelligence Act, whose transparency obligations became enforceable this month, according to the European Commission.
The rules cover four scenarios: AI systems that interact directly with people, AI-generated or manipulated "deepfake" content, synthetic text published for public information, and systems using emotion recognition or biometric categorization. Providers must disclose when content is AI-made and embed machine-readable markers so it can be detected later, unless the AI nature of an interaction is already obvious — an exception EU guidance interprets narrowly, according to law firm Cooley. Companies that fail to comply face fines of up to €15 million or 3% of global annual turnover, whichever is higher.
The obligations apply immediately to systems already on the market, though a transitional window runs until Dec. 2 specifically for the technical marking and detection requirements. More than 180 organizations, including major AI developers, have already signed onto a related EU Code of Practice on transparency for AI-generated content, the Commission said.
[Transparency obligations] reduce deception and manipulation and help people make informed choices. They also give businesses clearer obligations and a practical way to show compliance.
European Commission
The move is the latest phase of a staggered rollout of the AI Act, which began restricting "unacceptable risk" uses in 2025 and will add further obligations for high-risk systems in 2027. Enforcement now falls to the Commission's AI Office working alongside national regulators, which have opened complaint channels for consumers and a separate whistleblower tool for reporting violations.
Because the rules apply to any AI system reaching EU users regardless of where a company is based, compliance teams at U.S. and Asian AI firms have spent recent weeks adding disclosure banners and watermarking tools, according to legal trackers of the rollout — an early test of the compliance costs likely to accompany later, stricter phases of the law.