- Published on
OpenAI, Anthropic Call for Global AI Safety Standards; US Government Rejects New Global Governance Framework
- Authors
- Name
- DeepHorus
Leaders from OpenAI, Anthropic, and Hugging Face jointly called at the United Nations for stronger international coordination in artificial intelligence, hoping that countries would establish mutually comparable standards for AI capability assessments, safety testing, and incident reporting. However, the Trump administration explicitly opposes establishing a new global AI governance architecture for this purpose, arguing that the pace of AI development and its potential risks do not provide sufficient grounds for restricting technological development or transferring regulatory authority to international mechanisms.
This disagreement shows that, as frontier AI capabilities advance rapidly, leading U.S. AI companies and the Trump administration are not wholly at odds over safety risks themselves; both sides acknowledge that more powerful models may introduce new risks. The real disagreement lies in how those risks should be translated into regulatory systems, and how much coordinating and governing authority international institutions should have.
OpenAI Seeks to Establish Common Risk Assessment and Incident Reporting Standards
OpenAI CEO Sam Altman said at the United Nations that, as AI capabilities grow, the most important decisions concerning this technology cannot be made solely by a small number of laboratories based in San Francisco. Countries need to establish shared assessment and safety frameworks for powerful AI systems.
Altman proposed establishing standards at both the national and international levels to measure AI system capabilities, assess associated risks, examine safety safeguards, and determine the extent to which humans still retain oversight and control over AI systems. He also advocated creating mechanisms for faster incident classification, reporting, and disclosure, enabling different countries to use a common language to compare evidence.
