EU treats agents as AI systems

The European Commission is treating artificial intelligence agents as artificial intelligence systems already covered by the European Union’s AI Act, not as a new legal category. (ai-act-service-desk.ec.europa.eu) That position now appears in the Commission’s AI Act Service Desk under the question, “How are AI agents addressed within the AI Act?”, alongside the law’s existing guidance for general-purpose models and high-risk systems. The Commission’s broader guidance program has been underway since at least December 4, 2025, when it said it was preparing practical implementation guidelines for 2026. (ai-act-service-desk.ec.europa.eu) (digital-strategy.ec.europa.eu) The legal baseline is already in force. Regulation (European Union) 2024/1689 entered into force on August 1, 2024, and the Commission says the Act will be fully applicable on August 2, 2026, with some provisions already applying earlier, including the ban on prohibited practices from February 2, 2025. (digital-strategy.ec.europa.eu) An agent is usually software that can plan steps, call tools, and act across several systems with limited human prompting. The Commission’s move means companies cannot label that architecture an “agent” and assume it sits outside rules built for providers and deployers of artificial intelligence systems and models. (ai-act-service-desk.ec.europa.eu) (digital-strategy.ec.europa.eu) The AI Act regulates by risk, not by buzzword. The Commission says the law sets four levels of risk for artificial intelligence systems, from banned uses to high-risk uses to lighter transparency duties for lower-risk tools. (digital-strategy.ec.europa.eu) That matters for teams building “agentic” products in customer service, hiring, finance, health, or public administration. If the agent performs a use case listed as high-risk or relies on a general-purpose model with separate duties, the existing obligations on documentation, oversight, and compliance still attach to the system. (digital-strategy.ec.europa.eu 1) (digital-strategy.ec.europa.eu 2) The Commission has been building tools around that approach. Its AI Act Service Desk now includes an AI Act Explorer and a beta Compliance Checker that the Commission says is designed to clarify obligations and requirements under the law. (ai-act-service-desk.ec.europa.eu 1) (ai-act-service-desk.ec.europa.eu 2) Outside the Commission, legal analysts had been warning that “agent” was becoming a loose marketing label faster than it was becoming a legal one. Economist & Jurist, which first highlighted the Commission’s clarification this week, described the move as a quiet answer to a fast-growing compliance question in 2026. (economistjurist.es) The practical effect is narrow but immediate: map what the system actually does, then apply the AI Act category that fits. In Brussels, “agent” is now being treated as a design pattern, not a regulatory exit. (ai-act-service-desk.ec.europa.eu)