Brussels Joins the Slowdown: Von der Leyen Puts 'Pace the Frontier' at the Heart of Her State of the Union

For four days, “pace the frontier” was a fight inside the AI industry. As of Wednesday, September 16, 2026, it is also the declared position of the European Union.

In her annual State of the Union address to the European Parliament in Strasbourg, European Commission President Ursula von der Leyen did something no other head of government has done since Anthropic CEO Dario Amodei published his essay “We Must Pace the Frontier” on September 12: she picked a side. And the side she picked is the slowdown.

What She Said

The key passage, quoted by Reuters and Euractiv, was unambiguous:

“CEOs of the most advanced companies tell us that it is time to slow down on the self-recursive models. To pace the frontier. If the people developing the technology are clear, then we should be too.”

And then the operational announcement:

“I will invite the main frontier labs for a discussion on how we can support ongoing industry efforts to pace the frontier.”

That single sentence converts what had been a CEO-level moral argument — Amodei’s essay, Altman’s endorsement, Musk’s backing — into a planned governmental convening. No date has been set, no participant list published, but “the main frontier labs” in current Brussels parlance means OpenAI, Anthropic, and Google, the three companies that have spent months quietly building a joint industry standards body for exactly this purpose, alongside Google DeepMind and xAI.

The framing is notable: von der Leyen did not say Europe will impose a slowdown. She said Europe will support ongoing industry efforts. That is deliberate. The EU’s leverage over where OpenAI or Anthropic sets their capability dial is close to zero — those decisions are made in San Francisco and London. What Brussels can offer is legitimacy, structure, and the AI Act’s enforcement machinery, which formally came into force for frontier models on August 2, 2026.

The Threat Picture She Cited

Von der Leyen grounded the endorsement in a security argument, not a philosophical one:

“As the frontier models become more capable, these risks have come more sharply into focus. Models being developed will allow hacking on a level we never thought possible. And they will soon be in the hands of adversaries who see the world very differently from us.”

She pointed to recent incidents of rogue AI agents — the summer’s running series of agent misbehavior episodes that pushed the labs toward self-restraint in the first place — and to warnings from developers themselves. The backdrop includes former Anthropic researcher Jacob Coxon’s viral resignation warning that labs are racing toward self-improving superintelligence without adequate safety, a warning he has been repeating on American television this week.

Why This Matters: The First Government Endorsement

Until today, the entire “pace the frontier” movement was intra-industry. Amodei proposed it; Altman agreed “we need to pace the frontier” and committed to independent evaluators with employee-like access; Musk endorsed cross-lab test harnesses. Meanwhile the White House moved the other way — President Trump dismissed the safety discussion as “a hoax” and urged speed, and Nvidia’s Jensen Huang told a Dreamforce audience to “run as fast as you can” in a direct rebuke to Amodei.

That left the slowdown camp with a legitimacy problem: the people calling for restraint were the people selling the models. Brussels just changed that calculus. When the Commission President of a 27-member bloc with the world’s first comprehensive AI law says the CEOs are right and convenes them to formalize it, “pace the frontier” stops being a competitive positioning exercise and starts becoming policy infrastructure.

It also hands the labs a shield they have wanted for months. Coordinated slowdowns among competitors raise antitrust questions; a framework blessed — or at least hosted — by a government regulator outside US jurisdiction blunts that risk. Beijing has already rejected the “pace the frontier” proposal as a Cold War freeze play, which means the EU’s endorsement deepens the geopolitical split over who controls the tempo of AI progress.

The EU’s Next Moves

Von der Leyen outlined a concrete cooperation agenda alongside the invitation: Europe will work with “like-minded partners such as Canada, the United Kingdom and others” on model evaluation, verification, and AI security. She argued the AI Act, adopted after years of negotiation, puts Europe “in the position to shape global efforts” on frontier risk — a claim that will be tested quickly, given that Brussels spent the past month scolding labs for filing incident reports that are “not just a tick box” exercise, and ordered detailed disclosures of lab security practices on August 26.

The AI piece sat inside a broader security-heavy speech: she called for a “European Security Council,” proposed a social media ban for children under 13 with no personal accounts under 15, and outlined a defense mechanism for hybrid attacks. But the AI passage is the one with global consequence — it is the moment the world’s largest regulatory bloc aligned with the people building the systems rather than the people demanding acceleration.

What to Watch

Three open questions follow from today’s speech. First, attendance: whether OpenAI, Anthropic, and Google actually send decision-makers to Brussels, and whether xAI — whose CEO publicly endorsed pacing but whose company is the loudest acceleration holdout culturally — is invited at all. Second, substance: whether “support ongoing industry efforts” evolves into anything enforceable, such as EU recognition of the labs’ emerging standards body, or stays a talking shop. Third, Washington’s reaction: the Trump administration has already shown it views European AI regulation as overreach, and a Brussels-hosted slowdown convening is precisely the kind of “regulatory wall” Bloomberg reported could box in smaller rivals while the big labs write the rules together.

One thing is settled, though. On Saturday it was an essay. By Monday it was an industry consensus. On Wednesday it became a line in the State of the Union of the European Union. The pace of the “pace the frontier” story itself has been the fastest-moving narrative in AI policy this month — and it now has a permanent institutional address in Brussels.