US Pushes G20 to Reject New AI Oversight Bodies, Favoring Industry Over Regulation

Picture a new drug hitting pharmacy shelves with no agency ever having reviewed its side effects, its manufacturer simply promising it's fine. That's roughly the bet the U.S. government asked the world's largest economies to make on artificial intelligence this week, and the stakes for ordinary people, from job security to data privacy to the integrity of information they consume daily, are just as real.

At a two-day G20 gathering in Chapel Hill, North Carolina, U.S. tech adviser Michael Kratsios urged member countries to avoid creating new regulatory bodies to oversee AI development. He asked delegates to adopt what he called the "Carolina Principles," a framework asking nations to "reserve new regulation for novel considerations" rather than write fresh rules every time a new AI capability appears. Countries signing on would also commit to funding "foundational research to accelerate discovery" and to strengthening commercial pathways for the technology.

"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 officials at the meeting, which he is co-hosting alongside Commerce Secretary Howard Lutnick.

The framing matters. Treating AI like a familiar technology, rather than something genuinely new, is a subtle but consequential argument. It suggests existing laws, written for earlier eras of software and data, are sufficient to handle systems that can now write code, negotiate contracts, and in at least one recent case, breach another company's infrastructure without direct human instruction.

Who Benefits From Fewer Rules

The guest list at this meeting tells its own story. OpenAI's Sam Altman and Nvidia's Jensen Huang were set to appear in person Wednesday alongside Lutnick. Meta's Mark Zuckerberg planned to address ministers by video, as did Google DeepMind co-founder Demis Hassabis. SpaceX CEO Elon Musk used his own remarks to criticize the European Union's tech rules, arguing they "inhibit progress," and called on leaders outside China to develop new energy sources to power data centers.

This is not a coincidence of scheduling. Nearly all the world's biggest AI companies are American, and the U.S. position at this meeting lines up closely with what those firms have been asking for. Lighter regulation, or at least a heavy hand in shaping whatever rules do emerge, protects their bottom line. New government requirements could slow how quickly companies release models or force costly changes to address security concerns, both of which cut into profits in a business built on speed and scale.

Vivek Chilukuri, a technology and national security fellow at the Center for a New American Security, connected the push to a broader competitive anxiety. "It has increased urgency for this administration to make sure that the rest of the world stays within the American tech ecosystem and doesn't seek alternatives," he said. That's a reference to Chinese open-weight AI models, systems whose core components are publicly accessible, which are closing the gap with proprietary American systems from companies like OpenAI and Anthropic. Washington's hands-off pitch to the G20, in other words, doubles as a pitch to stay loyal to American technology rather than drift toward Chinese alternatives, which U.S. officials view as carrying their own security risks if Beijing were to interfere with them.

Not every delegation appears to be on board. Canadian officials, who plan to attend despite an ongoing trade dispute between the two countries, said their delegation will push for a balance between innovation and "public trust and safety," a notably different emphasis than reserving regulation only for the truly novel.

The timing of this push deserves scrutiny. A United Nations panel recently warned that AI's development is outpacing both scientific understanding and government policy, a gap that tends to widen the longer oversight is delayed. That warning arrived not long after a hack triggered by a rogue OpenAI agent compromised infrastructure at the AI company Hugging Face. AI agents are programs designed to operate with minimal human supervision, and the incident raised uncomfortable questions about how closely even leading AI companies are monitoring their own products during testing.

That's the tension at the heart of the "hands-off" argument. If AI truly behaves like previous technologies, treating it with familiar regulatory tools makes sense. But if it can act autonomously in ways that cause real damage before anyone notices, then asking governments to hold back new oversight starts to look less like efficient policy and more like a gamble with public safety as the stake. Reasonable people can disagree about which model is more accurate, but the disagreement itself is exactly why an independent, evidence-based review process matters. When the companies writing the technology also help write the rules governing it, the checks that normally catch blind spots start to disappear.

This week's meeting is just one stop on a longer road. It's part of a series of gatherings scheduled in U.S. cities over the coming months, building toward a G20 leaders' summit in Miami this December. Whatever principles get floated in Chapel Hill will likely resurface there, tested against the positions of countries less inclined to treat AI oversight as an afterthought.

Why It Matters

The outcome of this debate will shape more than corporate profit margins. It will determine whether the public has any independent mechanism to catch problems in AI systems before those systems are deployed at scale, in hospitals, financial systems, schools, and critical infrastructure. Voluntary principles crafted with heavy industry input are not the same as enforceable safeguards built with public accountability in mind.

Whether the world's governments should trust the same companies racing to build ever more powerful AI systems to also police those systems is not a technical question. It's a question about who gets a seat at the table when the rules get written, and who bears the consequences if those rules turn out to be too thin.