EU clarifies AI agents are covered

The European Commission’s latest Artificial Intelligence Act guidance is being read in Brussels as a sign that “artificial intelligence agents” are not outside the law just because they act autonomously. The Act already covers machine-based systems that generate predictions, content, recommendations, or decisions with some level of autonomy. (digital-strategy.ec.europa.eu) (eur-lex.europa.eu) The Commission published its guidance for general-purpose artificial intelligence models on July 18, 2025, ahead of obligations that started applying on August 2, 2025. The document says it is meant to give “legal certainty” across the artificial intelligence value chain. (digital-strategy.ec.europa.eu 1) (digital-strategy.ec.europa.eu 2) Under the Artificial Intelligence Act, the key legal question is not whether a product is marketed as an “agent.” The question is whether it is an artificial intelligence system, a general-purpose model, or a high-risk use case such as employment, education, critical infrastructure, or access to essential public and private services. (eur-lex.europa.eu) (digital-strategy.ec.europa.eu) That matters because high-risk systems face concrete compliance duties, including risk management, technical documentation, record-keeping, human oversight, and conformity assessment. The Act also imposes transparency and risk-management duties on providers of the most capable general-purpose models, especially those that may pose systemic risks. (eur-lex.europa.eu) (digital-strategy.ec.europa.eu) The timetable is staggered rather than one big launch. The ban on prohibited practices, the Act’s definitions, and artificial intelligence literacy rules have applied since February 2, 2025, while some governance, penalty, and general-purpose model obligations began on August 2, 2025. (eur-lex.europa.eu) The public-sector side of the story moved this month. On April 1, 2026, the European Union’s Interoperable Europe portal said the Commission would add a Public Sector Artificial Intelligence and Interoperability Readiness Pathway to its toolbox for administrations. (interoperable-europe.ec.europa.eu) That pathway is designed as a step-by-step guide for public bodies building artificial intelligence-enabled services. The portal says it is meant to turn “complexity into actionable steps” and help administrations design systems that can work across borders and with existing government infrastructure. (interoperable-europe.ec.europa.eu) The Commission’s broader line has been consistent: most artificial intelligence uses face limited rules, but systems that affect rights, safety, or access to services get tighter scrutiny. For companies selling “agentic” software into Europe, that leaves less room to argue that a new label changes the compliance burden. (digital-strategy.ec.europa.eu) (eur-lex.europa.eu)