The European Union’s technology chief, Henna Virkkunen, said the bloc’s artificial intelligence law already gives Europe the tools to deal with AI agents that escape human control, Reuters reported Friday. Her remarks show Brussels intends to answer recent safety scares with the binding rules it already has, not new voluntary pledges. More than 30 AI companies are already being questioned under those rules.

What Happened

In an interview with Reuters in Brussels, Virkkunen said the safety and security of very capable AI models is a pressing international issue and that Europe is well prepared for it. She said the AI Act “covers the whole life cycle of these models,” from development through deployment and ongoing use.

Reuters reported that regulators are advising companies on how to evaluate their models and how much time to set aside for that testing. Regulators will also weigh recommendations from a scientific panel of 60 AI experts drawn from universities and institutions.

RelatedWhite House AI Task Force Plans Breach Warnings From Labs at Its First Meeting

Virkkunen rejected complaints from some companies that the rules are already out of date. She argued that lawmakers planned for future advances by requiring companies to assess risks, bring in outside experts and keep monitoring their models after release.

She also said the European Commission is assessing replies to requests for information it sent in late August to more than 30 AI companies about their safety and security measures. Follow-up questions covered transparency and copyright. The Commission has not named the companies. It also sent requests to Chinese AI startups, and Virkkunen noted that the most capable models on the market come from the U.S. and China. According to Reuters, such requests can lead to formal investigations and fines of up to 7% of a company’s global annual turnover.

Her comments came after recent incidents at OpenAI and Anthropic that, Reuters said, have added to fears worldwide that AI systems could slip beyond human control. AI agents are systems that can take actions on their own, such as browsing websites or running software, with limited human supervision.

Asked whether the EU should impose a digital levy that would fall mainly on large U.S. technology companies, Virkkunen said that decision rests with EU governments. They are discussing new sources of money for the bloc’s 2028–2034 budget.

Why It Matters

The AI Act entered into force Aug. 1, 2024, and sets obligations according to how much risk an AI system poses. According to the European Commission, rules for general-purpose AI models have applied since Aug. 2, 2025. These are large models, like those behind chatbots and agents, that can be adapted to many tasks. Most of the law became applicable on Aug. 2, 2026. The Commission’s AI Office can now request technical documentation from model providers, evaluate models, require corrective measures and impose fines.

That puts the EU on a different track from Washington. In the U.S., six AI and chip companies signed a voluntary White House accord on Sept. 29 that relies on their own controls and audits and carries no legal penalties. Any developer selling advanced models in Europe answers to an enforcer with investigative and fining powers, whatever its home government asks of it.

For executives, the practical effect runs through procurement and compliance. Companies building on general-purpose models in Europe depend on their suppliers meeting EU testing and monitoring duties. The Commission’s review of the 30-plus responses is the first visible test of how hard the AI Office will press.

What to Watch

  • Whether the Commission’s review of the responses leads to formal investigations, and whether it names the companies involved.
  • The evaluation guidance itself, including how much testing time regulators expect and how far it follows the scientific panel’s advice.
  • Member states’ talks on financing the 2028–2034 budget, where a digital levy aimed mainly at large U.S. tech firms remains under discussion.