The European Union said it can respond to the risks posed by high-performance AI models that could slip into an uncontrollable state using its current AI regulations.

Cointelegraph, a blockchain media outlet, reported on Thursday that EU tech chief Henna Virkkunen said the EU’s AI Act could cover the risk of so-called “rogue AI agents.”

Virkkunen told Reuters in an interview that safety and security issues surrounding high-performance models have become a hot international issue and that Europe is well prepared. The comments come as concerns have grown about AI agents escaping human control after incidents involving OpenAI and Anthropic.

The EU’s AI Act sets different levels of regulatory intensity depending on risk. It was adopted 2 years ago and is considered one of the world’s strictest AI regulations. Virkkunen dismissed criticism that European rules have already fallen behind. She said the AI Act covers the “entire life cycle” of these models and that regulators are also providing companies with guidance on how to evaluate models.

That process reflected recommendations from 60 AI experts. Regulators are guiding companies on what standards to use to check a model’s risks and safety measures. It is understood as a move to examine not only the launch stage but the entire span from development to operation.

The European Commission in late August asked more than 30 unnamed AI companies to submit information. The request covered details of each company’s safety and security measures. Virkkunen said she is currently reviewing the responses. The EU has thus entered a process of checking how companies are responding at the enforcement stage.

Warnings have also continued within the AI industry about development speed and controllability. Dario Amodei (다리오 아모데이), chief executive of Anthropic, wrote in a blog post on Sept. 13 that AI development is moving too fast. He warned that if this trend is not controlled, it could “outpace our ability to understand and control this system.”

Amodei also argued that development is accelerating further as AI increasingly gains the ability to create the next generation of AI. He said recursive self-improvement is driving AI advances. That concern also aligns with why EU regulators are moving quickly to verify the safety of high-performance models.

Against this backdrop, the EU has reaffirmed the scope of its existing legal framework while entering a stage of gauging the level of follow-up action based on materials submitted by companies. Key points to watch are how the Commission will assess the replies from about 30 companies and whether it will issue additional guidance or enforcement measures for high-risk AI models.