EU Says AI Act Can Address Risks From Autonomous Agents

European Union technology chief Henna Virkkunen has defended the bloc's existing AI regulatory framework, saying it is equipped to address risks posed by increasingly autonomous artificial intelligence systems. Her comments, reported by Reuters on October 9, come as policymakers and researchers debate how to govern AI agents that can take actions with limited human intervention. The EU AI Act uses a risk-based approach to regulating artificial intelligence. Its obligations differ according to the type of system and the potential harm associated with its use. High-risk applications face stricter requirements, including oversight and documentation, while some uses are prohibited. Virkkunen argued that the framework was designed to account for future AI developments rather than only the capabilities available when the law was adopted. This matters because agentic AI systems can interact with websites, software and real-world processes in ways that go beyond generating text or images. The European Commission has also sought information from more than 30 AI companies worldwide about safety and compliance, according to Reuters. Such inquiries help regulators assess how firms identify risks, test systems and implement safeguards. Requests for information do not, by themselves, establish wrongdoing. The central policy challenge is how to balance innovation with accountability. Businesses want to use autonomous agents to reduce repetitive work and improve productivity, while regulators are concerned about unauthorized actions, insufficient human supervision and consequences when automated decisions go wrong. For developers building AI-powered products for European users, the discussion reinforces the importance of documenting system behavior, evaluating foreseeable risks and designing clear controls for human intervention. Companies should also distinguish between broad political statements and the specific legal obligations that apply to their products. The EU's position does not mean every emerging risk has been resolved. It signals that regulators intend to apply and interpret the existing framework as technology evolves, rather than assuming that every new class of AI system requires an entirely new law. As AI agents gain more operational authority, questions about monitoring, transparency and responsibility will remain central to both product design and regulatory enforcement.