The Pulse
U.S. Wins G20 Backing for Light-Touch AI Rules
The United States persuaded G20 ministers to endorse the Carolina Principles, a framework that limits new AI-specific regulation and favors targeted rules. The agreement puts Washington’s innovation-first approach against calls for stronger
AI.info Team ·
The United States entered a G20 technology meeting urging governments not to build new AI regulators. One day later, ministers endorsed the Carolina Principles, giving Washington a diplomatic win for a light-touch approach even as other countries pressed for stronger public safeguards.
The agreement came at a two-day Innovation Ministerial in Chapel Hill, North Carolina, hosted by the U.S. Department of Commerce and the White House Office of Science and Technology Policy. The framework calls on governments to reserve new rules for genuinely new problems, invest in foundational research and expand commercial adoption of emerging technologies.
White House officials presented the principles as an alternative to broad, technology-specific regulation. Canadian officials attending the meeting said their delegation would seek a balance between innovation and “public trust and safety,” according to Reuters.
Washington’s proposed limit on new AI agencies
Michael Kratsios, director of the White House Office of Science and Technology Policy, told G20 officials that governments should not treat every new technology as a separate policy problem. His proposal asked countries to “reserve new regulation for novel considerations” rather than create blanket rules or new institutions for AI oversight.
“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 said, according to Reuters.
The U.S. position closely tracks the interests of American AI companies, which are seeking room to release new models and services without a patchwork of national requirements. Reuters reported that the meeting’s industry participants included Elon Musk, while OpenAI Chief Executive Sam Altman and Nvidia Chief Executive Jensen Huang were scheduled to address ministers alongside Commerce Secretary Howard Lutnick.
Canada raises public-trust concerns
The dispute is not over whether AI needs rules, but over where those rules should come from and how quickly they should be written. Washington argues that existing sector-specific laws can handle many uses of AI and that new requirements should be limited to cases that cannot be addressed through current frameworks.
Canada’s delegation took a more guarded position. A Canadian government spokesperson told Reuters that officials would advocate balancing innovation with public trust and safety, a formulation that leaves room for oversight in high-impact areas without endorsing a single global regulatory model.
The meeting also took place against warnings that technical progress is moving faster than government understanding. Reuters cited a recent United Nations panel warning that AI development is outpacing scientific knowledge and public policy, while a cyber incident involving an AI agent and the infrastructure of Hugging Face added to concerns about how companies test and monitor autonomous systems.
China gives the agreement a strategic edge
Competition with China gave the U.S. proposal a second purpose beyond regulation. Chinese open-weight models, whose core elements are publicly accessible, are gaining ground against proprietary systems developed by American companies such as OpenAI and Anthropic.
Vivek Chilukuri, a technology and national security fellow at the Center for a New American Security, said the advance of Chinese models has increased the administration’s urgency to keep other countries inside the American technology ecosystem rather than allow them to seek alternatives.
The Chinese Embassy in Washington did not respond to Reuters’ request for comment about China’s participation in the North Carolina meeting. The White House later said the G20 ministers had reached consensus on the Carolina Principles, including policy frameworks designed to support innovation, workforce development, intellectual-property rules for AI and international standards.
The principles leave national control intact
The agreement does not create a binding global AI law. The published principles say G20 members should preserve national sovereignty in governing emerging technologies and adapt their approaches as technical capabilities and evidence change.
That language gives governments room to apply the framework differently. It also avoids the type of centralized AI authority that U.S. officials opposed at the Chapel Hill meeting, while encouraging voluntary cooperation, information sharing and partnerships between governments, companies and research institutions.
Howard Lutnick, the U.S. secretary of commerce, described the consensus as a signal to investors and technology companies. “The statement released by the G20 member ministers today gives entrepreneurs the confidence to invest in their ideas,” Lutnick said in a White House announcement.
The next test will come outside the meeting room, as governments decide whether the principles change their domestic rules. For now, the G20 has endorsed Washington’s preferred starting point: targeted regulation for new risks, existing laws for familiar sectors and no new AI bureaucracy by default.
Reuters’ original report described the U.S. push at the opening of the Chapel Hill meeting on September 1, 2026.