The US will request that the G20 refrain from establishing any governance structures for AI.

      The United States is urging the G20 not to create new institutions for managing artificial intelligence. A White House representative articulated this stance ahead of a two-day meeting that began in North Carolina on Tuesday, according to Reuters.

      The argument is more about procedure than substance, which lends it effectiveness. Washington is not advocating for specific regulations it favors; instead, it is urging member nations not to form entities that could develop any regulations at all.

      Commerce Secretary Howard Lutnick and White House technology advisor Michael Kratsios are hosting the meeting. This session is the first in a series leading up to the G20 leaders' summit in Miami in December, meaning that the position articulated this week will be open for negotiation among nations for the next three months.

      The guest list highlights the administration’s perspective on the issue. Sam Altman and Jensen Huang are participating in person, while Elon Musk is joining via video, which is an atypical mix for an intergovernmental gathering.

      Canada has indicated it will advocate for a different approach, seeking a balance between innovation and public safety and trust. It remains unclear whether China's embassy will participate.

      The G20 lacks the authority to force anyone to comply, which has implications both ways. It cannot establish a regulatory body, and any language it agrees on still serves as a reference point for national governments when they take action domestically.

      The administration conveyed a similar argument at the Paris AI summit, where JD Vance cautioned against hindering technological progress with regulation, and the U.S. chose not to endorse an international declaration regarding regulated AI usage.

      What has shifted is the context and importance of the discussion. A G20 process creates communiqué language that influences what national regulators think they are authorized to do, and an accepted statement regarding international oversight is more challenging to revise than a speech given at a summit.

      Europe arrives at this meeting having already implemented what the U.S. opposes. The AI Act is in effect, the Digital Services Act is being enforced against platforms like ChatGPT, and the Cyber Resilience Act started this month.

      The timing also poses challenges. This week, the chair of the Financial Stability Board informed G20 finance ministers that AI-related cyber risks represent the most pressing threat to financial stability, and that warning came through precisely the type of multilateral body that the U.S. wishes to remain uninvolved.

      A United Nations panel has also cautioned that AI development is outpacing both scientific understanding and government policies. This serves as a case for the establishment of new institutions, raised by an existing one.

      For European regulators, the immediate question is narrower than the overarching principle. Nothing agreed upon in Miami would eliminate the AI Act, but a G20 agreement against international oversight would make European enforcement appear more like an exception rather than a model.

      The counterargument warrants thorough articulation. Institutions created now would be based on a technology that is not well understood, staffed by individuals appointed for a context that will inevitably shift, and would be slow to adapt once established.

      Additionally, there is a competitive aspect that participants at the meeting may avoid explicitly addressing. A nation that houses most of the leading model developers stands to gain less from international oversight than nations that host none.

      Developing economies have different, and quieter, interests. Countries that do not build models or effectively regulate them stand to benefit the most from international standards but have the least ability to demand them.

      Whether this position endures depends on who is willing to contest it. The G20 operates by consensus, and a communiqué that remains silent on AI governance would bolster the argument that no governance is necessary.

      Miami will host the summit in December. In the meantime, the question is whether any member will be ready to invest the diplomatic effort needed to formulate a statement that the United States has declared it will not accept.