何が起こったのか
The Straits Times, republishing a Reuters report, says the U.S. urged G20 officials in North Carolina to avoid broad new AI regulation and instead focus on novel situations involving AI. Technology adviser Michael Kratsios promoted the Carolina Principles, while China reportedly signed them.
The Straits Times, citing Reuters, reported that the United States pressed G20 members on Sept. 1 to take a hands-off approach to AI regulation during a two-day technology and commerce meeting in Chapel Hill, North Carolina. Michael Kratsios, a U.S. technology adviser and co-host of the meeting, promoted what he called the “Carolina Principles.” According to the report, countries that signed the principles agreed not to create entirely new regulations for AI and instead to address novel situations involving the technology. Kratsios argued that policymakers should not treat every emerging technology as a separate, first-of-its-kind policy problem.
The report said Kratsios told reporters that China had signed the principles. It also said he did not provide a copy of the document and that the Chinese embassy did not immediately respond to a request for comment. China’s technology and science minister, Yin Hejun, represented the Chinese Communist Party at the gathering, according to the report. Kratsios described his bilateral meeting with his Chinese counterpart as “great,” but the article did not provide a joint statement, the principles’ wording, a list of signatories or evidence showing how the reported agreement was reached. The reported Chinese signature is therefore not independently confirmed in the supplied material.
Several technology executives addressed the meeting by video, according to the Straits Times report. Google DeepMind chief Demis Hassabis called for safety tests for AI systems. Meta chief Mark Zuckerberg argued that countries should not restrict open- models, whose core elements are publicly accessible. SpaceX chief executive Elon Musk criticised European Union technology regulation and urged countries outside China to develop new energy sources for data centres. Anthropic co-founder Tom Brown was scheduled to address the group on Sept. 2, while Reuters had previously reported that OpenAI chief executive Sam Altman and Nvidia chief executive Jensen Huang would separately appear with Commerce Secretary Howard Lutnick.
なぜそれが重要なのか
The meeting exposed competing approaches to : the U.S. and industry representatives argued against broad restrictions, while Google DeepMind’s Demis Hassabis called for safety testing. The report also links the debate to competition with China and the future of open- models.
The meeting matters because it could influence how major economies describe the policy problem posed by increasingly capable AI systems. The approach reported by the Straits Times would discourage broad, technology-specific regulation and reserve new rules for situations judged to be genuinely novel. That is a policy direction, not a demonstrated change in law. The source does not establish that the Carolina Principles are binding, that they have been formally adopted by the G20, or that they would prevent existing consumer-protection, competition, safety or sector-specific rules from applying to AI.
The discussion also revealed a divide over the balance between openness, innovation and safeguards. Hassabis’s call for safety tests points toward more systematic evaluation of AI systems. Zuckerberg’s opposition to restrictions on open- models reflects a concern that controls could limit public access to model components, according to the report. Musk’s criticism of EU rules represents a separate argument about regulatory burden. These positions were statements by individual industry leaders at the meeting; the source does not show that they constitute a shared G20 position or a final U.S. government policy.
The Straits Times report places the regulatory debate within a broader U.S.-China technology competition. It says Chinese open- models are becoming more capable relative to proprietary systems from U.S. companies and could attract users or businesses seeking alternatives to American platforms. Vivek Chilukuri of the Center for a New American Security told the outlet that the administration wants the rest of the world to remain within the American technology ecosystem. That makes the regulatory proposal relevant not only to safety and oversight, but also to market access, geopolitical influence and the distribution of AI infrastructure. The report does not independently establish the comparative capability of the models or quantify any resulting economic effect.
インタラクティブなメカニズム: 実際にどのように機能するか
この開発の背後にある基盤となるテクノロジーをインタラクティブに探索します。
Why can ethical evaluation not be reduced to one model score?
次に見るべきもの
The principles’ text, signatories and legal or political effect remain unclear. Further appearances by OpenAI’s Sam Altman, Nvidia’s Jensen Huang and Anthropic’s Tom Brown, as well as preparations for the December G20 leaders’ summit, may show whether the proposal becomes a broader policy position.
The first priority is verification of the Carolina Principles themselves. A publicly available text would clarify whether they are voluntary principles, a political declaration or something more formal; which countries signed; what counts as a novel situation; and how the approach relates to existing AI laws. The supplied report says no copy was provided and records no immediate response from the Chinese embassy. Until those gaps are addressed, the reported Chinese signature and the principles’ practical meaning should be treated cautiously.
The next phase of the meeting may provide more detail. The report says Anthropic co-founder Tom Brown was scheduled to speak on Sept. 2 and that OpenAI’s Sam Altman and Nvidia’s Jensen Huang were expected to appear separately with Commerce Secretary Howard Lutnick. Their remarks could indicate whether companies support common safety tests, limits on open- models, or a voluntary framework. Any such remarks would remain advocacy or policy positioning unless followed by a documented government agreement or concrete implementation.
The report also says the North Carolina gathering is one of a series of meetings leading to the G20 leaders’ summit in Miami in December. Coverage should track whether the principles appear in later communiqués, whether participating governments announce specific measures, and whether safety testing becomes an operational commitment. The article identifies data-centre development as a growing U.S. political issue and notes that protesters gathered outside the meeting. It does not establish the scale of public opposition, the environmental effects of any specific project or the outcome of the wider regulatory debate. Those remain meaningful unknowns.