ANALYSIS The unanimous adoption of the Carolina Principles by all G20 nations gives the Trump administration's light-touch AI governance model an international imprimatur, even as the very companies that stand to benefit most shaped the message from the podium. The consensus, though non-binding, reframes the global regulatory default away from comprehensive AI legislation and toward sector-specific rules, a posture that favors incumbent American labs over governments still drafting their own frameworks.
Why it matters
The Carolina Principles, endorsed at a two-day G20 Innovation Ministerial in Chapel Hill, North Carolina, hosted by the Commerce Department and the White House Office of Science and Technology Policy, call on governments to apply "existing sector-specific regulatory approaches to emerging technologies, where appropriate" and reserve new regulation for genuinely novel situations4. All G20 nations signed on unanimously2,5. The framework asks countries to avoid creating new regulatory bodies for AI governance entirely. ◆ That language effectively asks the rest of the world to adopt the posture the U.S. has already chosen domestically: no omnibus AI statute, no standalone regulator, and maximum room for commercial deployment.
The big picture
The Chapel Hill gathering doubled as a showcase for American AI power. Google DeepMind's Demis Hassabis, Meta's Mark Zuckerberg, and SpaceX CEO Elon Musk addressed delegates by video on Tuesday13. Nvidia CEO Jensen Huang and OpenAI CEO Sam Altman appeared in person alongside Commerce Secretary Howard Lutnick on Wednesday9. Anthropic co-founder Tom Brown was also slated to address the group. The lineup placed the executives who build frontier models and the chips that power them directly across the table from the ministers who write the rules.
Their messages tracked the administration's line. Huang urged G20 countries not to craft AI regulations focused on "theoretical harms" and instead focus on rules dealing with "real-world problems"7. Zuckerberg warned against placing restrictions on open-weight AI models. Hassabis struck a different note, calling on G20 officials to set up safety tests for AI systems. Musk urged leaders outside China to develop new energy sources to power data centers.
White House science and technology adviser Michael Kratsios, who co-hosted the meeting, framed the pitch in pragmatic terms: "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," he said. Lutnick acknowledged the diplomatic difficulty, noting that getting all participating countries to agree was "an enormous amount of work. When you bring global leaders together, there's always lots of politics".
Between the lines
China's presence at the table complicates any simple reading of the consensus. Chinese Science and Technology Minister Yin Hejun attended the sessions, and Kratsios described their bilateral meeting as "great"6. ANALYSIS Beijing's willingness to sign a U.S.-drafted deregulatory framework is notable given the broader rivalry: the principles impose no binding constraints on China's own AI development while aligning, at least on paper, with a posture that limits regulatory friction for Chinese firms operating internationally.
Meanwhile, the call for U.S.-China AI safety cooperation is gaining new advocates within the American AI establishment itself. Dean Ball, OpenAI's head of strategic futures, wrote on X that he had "always believed US/China AI safety collaboration would be desirable but thought it was unlikely to happen"1. ◆ That statement, arriving just as Chinese state media attacked American frontier governance, illustrates the gap between the diplomatic choreography at Chapel Hill and the unresolved question of whether safety norms can travel across geopolitical fault lines.
Australia's assistant technology minister, Andrew Charlton, argued that measures such as protecting original work from large language models and requiring data-center developers to supply their own energy are "not at odds" with the Trump administration's hands-off call15. ◆ The claim suggests that at least some signatories view the non-binding principles as compatible with domestic rules that go further than the Carolina text itself envisions.
A United Nations panel recently warned that AI developments are outpacing scientific understanding and government policy. ◆ The Carolina Principles offer no mechanism to close that gap; they instead codify the position that existing regulatory architecture is sufficient until proven otherwise.
What's next
The Chapel Hill ministerial is one of a series of G20 gatherings in U.S. cities leading up to the G20 leaders' summit in Miami in December. ◆ Whether the Carolina Principles survive contact with that higher-level forum, where heads of state face domestic political pressures that innovation ministers do not, will determine whether the consensus holds or fractures into competing regional regimes. For now, the U.S. has set the terms: the burden of proof falls on those who want new rules, not on those who want to ship models.