The United States plans to urge G20 participants gathering in North Carolina to avoid creating new artificial-intelligence-specific rules and regulatory bodies, according to Reuters reporting on prepared remarks by White House science adviser Michael Kratsios. The proposed approach, called the Carolina Principles, would extend the administration’s preference for a minimally burdensome, sector-based AI policy into an international forum.
The two-day meeting, scheduled to begin September 1 and co-hosted by Commerce Secretary Howard Lutnick and Kratsios, brings commerce ministers together with technology executives ahead of the G20 leaders’ summit expected in Miami in December. It is not itself a G20 agreement, and no public text of the Carolina Principles or evidence of other countries accepting them has been released.
What the proposed principles would do
In the Reuters account carried by MarketScreener, Kratsios was expected to ask countries to reserve new regulation for novel considerations, while investing in foundational research and expanding commercial opportunities for emerging technologies. The formulation is more limited than a claim that AI needs no rules: it argues against treating every new technology as a wholly new policy category.
The available reporting does not specify which issues would qualify as novel, how signatories would apply the test, or whether participants would be asked to endorse common language. Countries could reach different outcomes under such a principle because their consumer, competition, privacy and safety laws differ.
Governments also differ over whether existing institutions are sufficient for AI. Reuters reported that Canada planned to stress a balance between innovation, public trust and safety. That position does not establish a direct rejection of the U.S. proposal, but it illustrates a policy tension at a meeting intended to link AI development with trade and industrial strategy.
A policy position already visible in Washington
The North Carolina initiative would not represent a sudden reversal in U.S. policy. On December 11, President Donald Trump issued an executive order on a national AI policy framework that described the federal objective as sustaining U.S. AI leadership through a minimally burdensome framework. It directed federal agencies to examine state laws that might conflict with that goal and called for recommendations on a uniform federal legislative framework.
The order is also a useful corrective to the shorthand of a purely hands-off policy. It contemplated a federal framework while identifying areas that its legislative recommendation should not preempt when state laws are otherwise lawful, including child safety, AI compute and data-center infrastructure, state procurement and government use of AI. The order therefore combines an argument for national consistency with recognition that some AI-related questions remain tied to specific uses and institutions.
Kratsios set out the international version of that view at the India AI Impact Summit in February. In official White House remarks, he rejected global AI governance and advocated rules tailored to use cases and sectors, with existing frameworks updated where necessary. A separate White House summit statement emphasized AI adoption, sovereignty and exports.
Read together, those documents establish the administration’s published direction: reject global AI governance, seek fewer conflicting domestic requirements, and connect AI policy to U.S. technology deployment abroad. The Carolina Principles, as described by Reuters, would provide a possible diplomatic vehicle for that agenda rather than a newly documented change in its underlying philosophy.
Commerce ministers, industry leaders and geopolitical stakes
The event’s design reflects how AI policy has become entwined with economic diplomacy. Reuters reported that Mark Zuckerberg was expected to participate by video and that Google DeepMind co-founder Demis Hassabis was also expected. The meeting places industry figures alongside officials responsible for commerce, rather than convening a standalone technical standards or safety body.
Reuters also reported that competition with China formed part of the backdrop, including an assessment from Vivek Chilukuri of the Center for a New American Security that advances in Chinese open-weight models make it more urgent for countries to remain in the U.S. technology ecosystem. That is an expert view reported by Reuters, not an announced G20 position.
It remains unclear whether the U.S. can translate its preference for narrowly targeted oversight into language other governments will support. For now, the most concrete public record remains the December executive order and Kratsios’s February speech. The proposed Carolina Principles remain an unreleased initiative reported from prepared remarks, with no announced signatories or adoption outcome.