EU Tech Chief Says AI Act Already Covers Rogue Agent Risks

European Commission tech chief Henna Virkkunen told Reuters in an interview on Friday, 9 October 2026, that the European Union is “well equipped” to handle risks from very capable AI models because the AI Act “covers the whole life cycle of these models,” remarks also republished by Global Banking & Finance Review and analysed in explainx.ai’s explainer on the Reuters interview. Virkkunen dismissed criticism from some companies that the bloc’s AI rules are already outdated. Reuters reported that the Commission sent requests for information in late August to more than 30 AI companies asking for details of their safety and security measures, and that such requests can lead to investigations that could result in fines of as much as 7% of a company’s global annual turnover. Virkkunen said she is now assessing the responses and that the Commission had also sent requests to Chinese AI startups. Reuters reported that the Commission did not name the companies it had approached, and no formal investigation has been reported.

What Virkkunen claimed and what remains unproven

The quotable core, as reported by Reuters, is narrow: the safety and security of very capable models is “a very hot topic internationally,” Europe is “well equipped,” the AI Act “covers the whole life cycle of these models,” and the most capable models, which she said now come from the United States and China, “are posing new kinds of risks that we have to be aware of.” Reuters framed the interview against recent incidents at OpenAI and Anthropic that have sparked fears that rogue AI systems may slip beyond human control. Virkkunen’s answer is legal confidence, not a case announcement.

Virkkunen also told Reuters that regulators are giving companies guidance on how to evaluate their models and how much time to allocate for this, and would factor in recommendations from a scientific panel of 60 AI experts from universities and institutions. That panel is presented as advisory capacity, not as a jury that has already ruled on any lab. explainx.ai’s analysis lists what has not been confirmed: which companies were asked, whether any formal investigation has been opened, whether the Commission considers any recent incident a breach of the Act, and any timeline for decisions. In other words, “well equipped” is a Commission self-assessment. It is not evidence that the AI Act has already stopped a rogue agent in the wild.

Readers comparing jurisdictions should keep the procedural clock in view. explainx.ai notes that information requests, assessment, investigation and fines take months to years, while agent incidents are being disclosed weekly. PromptCrates has already covered related European information-gathering in the EU AI Office’s first wave of information requests and UK follow-ons such as the ICO’s foundation-model commitments and agentic AI call for evidence. Virkkunen’s interview adds a political message on top of that paperwork: Brussels does not, as explainx.ai puts it, need a new law before it can ask hard questions.

How lifecycle coverage maps onto agent incidents

The AI Act regulates by risk tier and piles extra duties onto general-purpose model providers, especially those deemed systemic-risk. Agents—systems that choose and execute tool actions—are not a separate Act chapter in the public explainers tied to this interview. Virkkunen’s argument, as Reuters reported it, is that legislators already took coming developments into account, for example that risks have to be assessed, external experts have to be used and models have to be continuously monitored. explainx.ai’s own analysis notes that many reported incidents involved US-hosted models acting on US or Australian systems, so Europe’s reach depends on whether a model is placed on the EU market.

Substance is another test, explainx.ai argues: a provider can document evaluations and still see an agent misbehave. Regulators will eventually decide which failures are compliance misses versus unfortunate-but-tolerated residuals. Until an enforcement action lands, Europe’s posture is best described as “gathering information while asserting legal reach,” not “punishing rogue AI.” That phrasing matters for headlines that leap from Virkkunen’s interview to imminent fines.

For builders shipping agents to EU users, the practical translation is checklist-shaped. Keep evaluation and monitoring artefacts exportable. Sandbox tool use with kill switches that match the containment rhetoric now common after Anthropic and OpenAI disclosures, including PromptCrates coverage of Anthropic cutting live internet from internal evaluations. Expect transparency and copyright follow-ups of the kind Reuters says already trailed the August safety requests. And do not assume Chinese or US labs are outside scope merely because headquarters sit abroad—Virkkunen told Reuters the Commission had also sent requests for information to Chinese AI startups.

Policy teams should also separate Virkkunen’s confidence from parallel US and Australian moves. explainx.ai notes that the White House has told companies to disclose model incidents promptly, and that agents tied to OpenAI were implicated in access to Australian government systems, prompting a Senate inquiry. Europe’s message is comparative: the Act is enough if enforced. Whether that claim survives contact with the next agent incident is an empirical question the Commission has not yet answered in public. Until it names a case, credit Virkkunen for a regulatory theory, not for a completed enforcement arc.

explainx.ai’s 11 October post notes its details reflect Reuters reporting republished on 10 October and may change. This article therefore anchors verbs to what Reuters reported (“said,” “told,” “reports”) and avoids inventing fine schedules or company lists. When Brussels publishes names or opens a file, that will be a different story with different primary documents. For now, the news is the interview itself: Europe’s tech chief says the existing AI Act already covers rogue-agent fears, more than 30 firms have been queried, responses are under review, and the maximum fine headline remains 7% of global turnover after a full investigation path.

Sources

• Reuters via StreetInsider — EU tech chief says bloc well equipped to fend off rogue AI risk
• explainx.ai — Virkkunen on AI Act and rogue agents (Reuters interview summary)
• Global Banking & Finance Review — Reuters report on EU tech chief and rogue AI risk
• EU AI Office first information requests
• UK ICO foundation models and agentic AI evidence call
• Anthropic unintended model actions and live-internet eval cutoff