On the first day of September, the United States and the European Union each made their AI-regulation strategy unmistakable — and the two could hardly be further apart.

In Chapel Hill, North Carolina, the US hosted a G20 "innovation" ministerial, led by Commerce Secretary Howard Lutnick and White House science and technology chief Michael Kratsios. The pitch to the assembled ministers was a light-touch doctrine Kratsios branded the "Carolina Principles": that regulation should not single out specific technologies. "Policymakers do not need to approach each innovation in isolation and should not treat every emerging technology as a first-of-its-kind policy problem," Kratsios told the meeting. The room reflected the message — guests at the conference included Nvidia's Jensen Huang, OpenAI's Sam Altman and Anthropic co-founder Tom Brown, with Meta's Mark Zuckerberg and Tesla's Elon Musk also reported present. This was the US making its case, with the AI industry's principals in the room, that AI needs no rules of its own — and the case landed: the ministerial closed with the G20 adopting the Carolina Principles as a consensus statement, Europe's own member states and China included.

Brussels chose the same day to demonstrate the opposite. The European Commission sent formal information requests to more than 30 AI companies worldwide, focused on safety and copyright compliance — a preliminary investigative step under the EU's AI Act. Henna Virkkunen, the Commission's executive vice-president for tech sovereignty, framed enforcement as non-negotiable: "Our goal is to ensure that AI in Europe is developed, released and used safely and transparently," she said, adding that Brussels was "ready to take all necessary steps" to enforce the Act.

Two doctrines, one day, and no overlap between them.

The disagreement is not really about whether AI is risky. It is about who should carry the cost of that risk, and when. The US position, as argued at Chapel Hill, is that technology-specific law is premature and clumsy — that existing rules on fraud, safety and competition already reach AI's harms, and that writing bespoke statutes for each new capability slows innovation without making anyone safer. The EU position is that waiting for harm to materialise before regulating is exactly the mistake it does not intend to repeat, and that a horizontal law with real enforcement teeth is the only way to make safety and transparency binding rather than voluntary.

For any company building AI at scale, the practical problem is that both regimes now apply at once. A model trained in California and served in Frankfurt answers to a US administration that wants no AI-specific obligations and a European Commission that is already sending it compliance demands. That is not a stable equilibrium a firm can optimise around; it is two moving targets. The information requests Brussels sent are a preview of the operational reality — safety documentation, copyright provenance, transparency reporting — that a US-headquartered lab must produce for the European market regardless of what Washington prefers.

This is the "Brussels effect" being tested in real time. For a decade, EU rules on privacy and product safety became de facto global standards because it was cheaper for multinationals to build to the strictest regime than to maintain two. The open question in AI is whether that still holds when the US is not merely declining to match the EU but is actively lobbying other G20 governments to reject the European approach. If enough countries adopt something like the Carolina Principles, the EU's rules stop being the global default and start being a regional tax — one that AI companies may increasingly try to route around rather than absorb.

What made the first of September notable was not any single announcement but the symmetry of it: the world's two largest regulatory powers, on the same day, each choosing to make their divergence explicit rather than paper over it. For the AI industry, the era of hoping the two would eventually converge on one rulebook looks to be over. The G20 has already signalled its lean — the Carolina Principles passed as a consensus, Europe's own member states included. The sharper question is the narrower one: whether the EU Commission's binding enforcement holds as the global outlier, or whether the "Brussels effect" can still pull the industry toward the stricter regime after the world's largest economies have endorsed the lighter one.