When the world's twenty largest economies sit down to talk about artificial intelligence, you might expect the host to arrive with a rulebook. Instead, at this week's G20 Innovation Ministerial in Chapel Hill, North Carolina, the United States showed up with the opposite: a pitch that nothing, not regulation, not power shortages, not even serious safety concerns, should be allowed to slow AI down. Michael Kratsios, who runs the White House science and technology office, told delegates that "slow policy adaptation makes it harder to capture the full benefits of innovation," and unveiled a framework the administration is branding the Carolina Principles: pro-innovation, industry-led standards, and pathways to commercialization rather than gates in front of it.
The most quotable line came from David Sacks, the administration's AI czar, who warned it would be a "disaster" to build an FDA for AI. "Any regulatory agency that has to approve AI product releases will become like a DMV for AI, where new models get stuck in a long queue waiting for approval," he said, arguing the technology is "just too dynamic and fast-moving for that type of approach." The message to the other G20 governments was that existing laws already cover any harm worth covering, and that a dedicated AI regulator would simply hand the lead to whichever country regulates least.
What makes this more than boilerplate is who the argument was aimed at. As CNET notes, the "DMV for AI" jab lands squarely on the one major AI company that skipped the event: Anthropic. Its chief executive, Dario Amodei, recently argued in an essay that frontier models, like airplanes, should undergo technical testing and auditing, and that releases should be blocked or reversed if they fail to meet a high safety bar. So the spectacle in Chapel Hill was, in part, a government using an international stage to rebut one of its own domestic firms. That is a striking inversion of the usual pattern, in which industry lobbies against rules and regulators push back.
The friction is not new. Anthropic and Washington have clashed before: a Pentagon deal collapsed and the company was briefly labelled a supply-chain risk before a judge intervened, and the government at one point ordered it to bar foreign nationals from using its Fable 5 model, prompting the firm to pull and later quietly re-release it. Set against that history, the G20 remarks read less like abstract policy and more like the latest round in a running argument over how much say the state gets in what these labs ship.
The hands-off camp had powerful backup. Elon Musk, appearing virtually, endorsed a "default legal as opposed to default illegal" philosophy for new inventions and argued governments over-protect incumbents at the expense of startups, a populist framing that sits awkwardly with his own status as the world's richest person and the recipient, per a Washington Post tally, of 38 billion dollars in government funding last year. Whatever is agreed in North Carolina feeds into the larger accord that G20 leaders are due to sign in Miami this December. The open question the forum leaves hanging is a simple one: if the United States exports a norm of minimal oversight, and the rest of the G20 signs on, who is left to catch the failures when a model does cause real harm?