EU Tech Chief Says AI Act Can Handle Rogue AI Risks Amid OpenAI, Anthropic Concerns

European Union tech chief Henna Virkkunen has declared that the bloc's regulatory framework is well-equipped to address risks posed by rogue AI agents, responding to growing global concerns sparked by recent incidents at OpenAI and Anthropic. In an interview with Reuters on Friday, Virkkunen emphasized that the safety and security of advanced AI models remains a priority for regulators worldwide, and that Europe is prepared to lead on this front.

The EU's AI Act, adopted two years ago and widely regarded as one of the world's strictest AI regulations, takes a risk-based approach to governing artificial intelligence development and deployment. Virkkunen pushed back against critics who argue the rules are already outdated, noting that the legislation covers the entire lifecycle of AI models. Regulators are actively issuing guidance to companies on model evaluation, drawing on recommendations from a panel of 60 AI experts assembled by the Commission.

The European Commission took a concrete step in late August when it requested detailed safety and security information from more than 30 unidentified AI companies, the responses to which Virkkunen is currently reviewing. The probe reflects heightened regulatory scrutiny as industry leaders warn of AI systems potentially outpacing human oversight. Anthropic CEO Dario Amodei wrote in a September 13 blog post that the pace of AI development may "outrun our ability to understand and control these systems," pointing to recursive self-improvement as a key driver of accelerating capability gains.

These warnings have intensified debate over whether AI progress should be temporarily slowed. Critics counter that any pause could destabilize economies already increasingly dependent on AI-driven growth. Virkkunen's reassurances signal that Brussels intends to enforce its existing framework rather than overhaul it, betting that structured oversight can keep pace with rapidly evolving frontier models.