EU's AI Enforcement Board Meets Thursday to Push Brussels' Rulebook as the Global Standard

Since the EU AI Act's enforcement provisions took effect on August 2, 2026, Brussels has been on a campaign to make its rulebook the global template. That effort reaches a new stage Thursday, September 17, when the Artificial Intelligence Board, the coordination body representing all 27 EU member states, convenes to discuss international AI policy, recent AI-related incidents, and a framework for regulating advanced models worldwide, according to Crypto Briefing.

The meeting isn't happening in a vacuum. On September 11, European Commission Executive Vice President Henna Virkkunen told a child-safety conference in Dublin that global AI rules are necessary, according to Politico Europe as reported by ua.news. Virkkunen argued the EU already requires companies including Anthropic and OpenAI to assess and mitigate the risk of losing control over their models, and that no comparable requirement exists anywhere else.

What's actually in force right now

Despite the headlines about Europe leading a global crackdown, most of the AI Act's toughest provisions aren't live yet. According to the European Commission's own AI Act page, the outright bans on practices like social scoring and certain biometric surveillance took effect back in February 2025. A ninth prohibited practice doesn't kick in until December 2026. And the strict pre-market obligations for high-risk AI systems, covering things like medical devices, hiring tools, and law enforcement applications, don't start until December 2, 2027, after amendments in the so-called AI Omnibus pushed the deadline back.

So the enforcement era that began August 2 is narrower than the political rhetoric suggests. It covers the clearest bans and general-purpose model transparency rules. The heavy compliance lift for high-risk systems is still more than a year away.

The pressure campaign for global buy-in

Virkkunen's push for international alignment comes against a backdrop of genuine alarm inside the AI industry itself. A former Anthropic researcher, Jacob Coxon, resigned and warned that people building AI systems believe the technology could kill everyone by the end of the decade, according to UN News. Anthropic and OpenAI have also disclosed that their models broke out of testing environments, incidents that UN High Commissioner for Human Rights Volker Türk cited in an open letter calling the current AI race a step change towards greater existential risks to every aspect of our lives.

Türk called on countries hosting frontier AI companies, which he identified as the U.S. and China, to require incident reporting, capability verification, and human rights due diligence, and to align their approaches so competition between the two doesn't turn into a race to the bottom.

Anthropic CEO Dario Amodei responded with his own essay, "We Must Pace the Frontier," arguing companies should keep advancing AI while building in third-party evaluation and cross-company coordination, according to UN News. OpenAI, Google, and xAI leadership reportedly signaled agreement with Amodei's framework, a notable moment of industry consensus on self-regulation rather than government mandates.

Türk welcomed that but said it isn't enough. "No company should be able to decide by itself which risks the world must accept," he wrote, adding that governments in states hosting frontier labs have "a particular obligation" to hold them accountable.

Washington isn't on board

The U.S. isn't rushing to copy the EU's approach. Andrew Puzder, the U.S. ambassador to the EU, said he supports an EU-U.S. AI partnership but explicitly warned against regulating American platforms and companies in a way that could weaken their ability to compete with China, according to ua.news. If European-style compliance mandates raise the cost of building and deploying frontier models for U.S. firms while Chinese labs face no equivalent constraint, the result is a competitive handicap for the side of the AI race that actually has to answer to voters and courts. Germany's own digital affairs minister, Carsten Wildberger, has floated a different model entirely: an international AI safety body along the lines of the International Atomic Energy Agency, rather than exporting the EU's domestic rulebook wholesale.

Irish MEP Regina Doherty has formally asked the European Commission whether it plans to push an international incident-response protocol at the G7, G20, and OECD. No answer has been made public yet. And European officials are watching to see whether the AI arms race comes up when President Trump and Chinese President Xi Jinping meet later this month, a conversation that will likely tell Brussels more about the odds of global alignment than Thursday's board meeting will.