What Brussels is doing differently

Read Virkkunen's Friday interview next to the week's US debate and the contrast is sharp. In Washington, the conversation is about whether labs will voluntarily slow down and whether they need an antitrust waiver to coordinate. In Brussels, the questions have already been sent. The Commission is not asking the industry to design an oversight body. It is using the one its law already created.

Virkkunen framed it bluntly: "The EU has the world's first law addressing the systemic risks that AI can pose." Where American policy chiefs talk about "moral" commitments, Europe's tech chief talks about RFIs, assessments, and fines. She expects the countries that have resisted AI regulation to "come back to that quite soon."

The part nobody else will say

The interesting line in this story is the one about Chinese startups receiving the same letters. Virkkunen confirmed it, noting that "some of the most capable models currently available" come from the United States and China. This is not a trade move wearing safety clothes. The questionnaires cover the same safety, transparency, and copyright ground for every recipient. The point is jurisdiction: if your model runs in Europe, Europe's questions find you, wherever you are incorporated.

There is also a second-order signal. The Commission did not send these in October after the latest incidents. It sent them in August, before the current wave of agent panic. The timing suggests the EU AI Office had its own threat picture and acted on it before the headlines caught up.

If you ship models into Europe

The RFI round was the warning shot; the assessment round is the sorting. Companies whose responses show documented safety evaluations, transparency measures, and copyright compliance will fade into the background. Companies whose responses are thin are now on a list for formal investigation. If you operate a model in Europe, the practical advice is old and suddenly urgent: document your evals, your safety practices, and your copyright posture now, because the Commission may ask you to produce them. The era of "we test internally" as an answer is ending in Brussels first.

Sources

  1. [1] Reuters — “EU tech chief says bloc 'well equipped' to fend off rogue AI risk” (Oct 9, 2026), via ETC JournalRead source
  2. [2] European Newsroom — “EU tech chief expects countries leading AI race to come around on safety” (Oct 9, 2026)Read source
  3. [3] Analytics Insight — “Can the EU AI Act Control Rogue AI? Brussels Defends its Regulatory Framework” (Oct 9, 2026)Read source
  4. [4] Weex — “EU Targets Risks of Autonomous AI, Inspects Safety Systems of Over 30 Companies” (Oct 9, 2026)Read source
  5. [5] PubAffairs Bruxelles — “Commission holds special meeting of Scientific panel on frontier AI safety and risks” (Oct 9, 2026)Read source