Europe’s digital policy chief insists the bloc already holds the legal tools needed to restrain dangerous autonomous AI systems, pushing back against fears that regulators remain a step behind the technology.
Henna Virkkunen, the EU’s top tech official, told Reuters on Friday that the bloc’s landmark AI Act provides sufficient safeguards against so-called rogue agents. She framed the law as a full-lifecycle framework, covering everything from model development through deployment and ongoing oversight.
That claim arrives amid mounting international anxiety over highly capable AI models. Virkkunen acknowledged the technology introduces genuine and unfamiliar hazards, yet she expressed confidence that Europe’s rulebook anticipates those challenges.
The European Commission fired off formal information requests in late August to more than 30 AI firms, demanding specifics about their safety protocols and risk controls. Virkkunen said she is now reviewing those responses. Regulators have also been issuing practical guidance to help companies evaluate their own systems.
Meanwhile, the law’s architect says policymakers did not overlook the need for external scrutiny. Virkkunen pointed to mandatory risk assessments, expert involvement and continuous monitoring requirements baked into the AI Act’s structure.
The statute, first agreed provisionally in December 2023, became the world’s first comprehensive and binding piece of AI legislation. It sorts foundation models and frontier systems into four risk tiers, each with escalating restrictions on market entry.
That framework drew criticism from industry voices and some government officials who argued the law prioritized danger mitigation over innovation. Supporters countered that Europe’s approach could become the global regulatory benchmark.
Recent events have sharpened the debate. Japan disclosed a scramble to strengthen cyber defenses after hacks hit over 20 major companies. Sophisticated AI-driven attacks, cheap and easy to deploy, reportedly worsened that threat. Two weeks earlier, the CEOs of Anthropic and OpenAI told the UN Security Council that governments and industry must cooperate on global AI guardrails.
