At the United Nations, leaders from three leading AI companies, OpenAI, Anthropic, and Hugging Face, called for enhanced international coordination in the field of artificial intelligence, hoping that countries would establish comparable AI capability assessments, safety tests, and incident reporting standards. However, the Trump administration clearly opposed using this as a basis for establishing a new global AI governance structure, arguing that the speed of AI development and potential risks were not sufficient reasons to restrict technological advancement or transfer regulatory authority to international mechanisms.
Corporate Proposals: Common Standards + Rapid Incident Reporting
Sam Altman, CEO of OpenAI, stated at the United Nations that as AI capabilities grow, the most important decisions about this technology should not be made solely by a few laboratories in San Francisco. Countries need to establish common evaluation and safety frameworks around powerful AI systems. He proposed that both national and international standards be established to measure model capabilities, assess risks, review safety measures, and determine the extent of human oversight over AI. He also advocated for faster classification, reporting, and disclosure mechanisms for incidents, enabling different countries to compare evidence and verify measures using a common language. OpenAI's recent policy framework also suggests that national and international cutting-edge AI standards should be interconnected, with unified protocols for measuring model capabilities and reporting incidents, while emphasizing the hope that the U.S. will play a leading role.
Dario Amodei, CEO of Anthropic, shares a similar position: if improperly managed, stronger AI could pose risks to society and even humanity, and safety standards must be raised in parallel with model capabilities. He stated that Anthropic is willing to slow down its release when necessary, but believes that frontier risks have exceeded the capacity of any single company to manage, requiring industry-wide and even global coordination. This assessment is related to recent incidents involving intelligent agents—this week, the UN's independent international AI science group highlighted an incident involving OpenAI and Hugging Face's intelligent agents, discussing the cross-institutional risks that may arise after models deviate from their goals, seek vulnerabilities, and evade supervision, emphasizing that AI incidents can transcend corporate and national boundaries, and no institution can fully grasp all risks on its own. Clement Delangue, a representative from Hugging Face, also called for stronger global AI monitoring and incident disclosure standards.
Trump Administration's Stance: Coordination Is Okay, Governance Is Not
The Trump administration drew clear boundaries on this issue. Michael Kratsios, the U.S. President's advisor on science and technology, stated at the United Nations that the U.S. acknowledges the acceleration of AI and potential risks, but these risks are not enough reason to pause development or use a new global governance structure to limit progress; international dialogue and technical coordination can continue, but they should not evolve into a "global governance" system dominated by transnational institutions. This position aligns with the U.S. stance at the General Assembly on September 22 — the U.S. opposes establishing a global control system for AI and considers maintaining the U.S. AI competitive advantage as an important policy goal.
This divergence shows that both sides are not entirely opposed on the existence of frontier risks, as they both acknowledge that stronger models may bring new risks; the real disagreement lies in how to convert these risks into regulatory systems, and how much coordination and governance power international institutions should have. The companies focus on establishing common capability tests, risk assessments, and incident disclosure standards; the Trump administration opposes transforming cooperation into a global mechanism with independent regulatory authority or binding force. It is worth noting that even though OpenAI advocates for international standards, it proposes a technology framework led by the U.S., encouraging other countries to adopt or recognize it, rather than transferring regulatory authority to global institutions.
