EU says agents are covered

The European Commission’s AI Act help desk now says artificial intelligence agents are covered by the law, closing a gap around systems that can act with some autonomy. (ai-act-service-desk.ec.europa.eu) The new question on the Commission’s AI Act Service Desk asks, “How are AI agents addressed within the AI Act?” and places the issue inside the law’s frequently asked questions for general-purpose artificial intelligence models. The Service Desk says it updates the list regularly based on stakeholder submissions and Artificial Intelligence Pact webinars. (ai-act-service-desk.ec.europa.eu) That matters because the Artificial Intelligence Act does not create a separate legal category for “agents.” The law instead regulates “artificial intelligence systems” and “general-purpose artificial intelligence models,” and the Commission’s February 6, 2025 guidance said coverage turns on whether a system can infer outputs with varying levels of autonomy. (paulweiss.com) (eur-lex.europa.eu) An agent, in plain terms, is software that does more than answer a prompt once. It can plan steps, call tools, and take actions across websites, files, or business systems with limited human input, which is why lawyers and companies had been asking where it fits under the Act. (iapp.org) (ai-act-service-desk.ec.europa.eu) The Commission is rolling out that answer through a centralized structure it began building in 2025. It launched the AI Act Service Desk and Single Information Platform on October 8, 2025, and said the tools were meant to give companies “legal certainty” as the law phases in. (publicnow.com) The enforcement model is not fully centralized, but Brussels has given itself a larger coordinating role than many earlier European digital rules. The European Commission says the European Artificial Intelligence Office oversees enforcement and implementation across member states, while national market surveillance authorities enforce rules for artificial intelligence systems at country level. (digital-strategy.ec.europa.eu) That split is already on the calendar. The Commission says each member state should have designated national competent authorities by August 2, 2025, while the main body of the Act applies from August 2, 2026, with some provisions applying later, including parts scheduled for August 2, 2027. (digital-strategy.ec.europa.eu) (eur-lex.europa.eu) The law itself entered into force on August 1, 2024, and the Commission has described it as a single-market rulebook for trustworthy artificial intelligence. Folding agents into that framework means companies building agentic products for Europe now have less room to argue that the systems sit outside the Act’s existing categories. (commission.europa.eu) (publicnow.com)