AI Executives Urge Global Safeguards at UN

A coalition of 22 governments and political leaders has called for globally consistent safeguards on frontier AI, including mandatory pre-deployment testing, independent evaluations, transparent safety reporting, shared reporting of serious incidents and common international standards. The proposal includes exploring a potential U.N.-backed institution to set rules and verify compliance when systems cross capability thresholds, while stopping short of calling for a pause in AI development. At a U.N. Security Council session, OpenAI, Anthropic and Hugging Face executives similarly urged urgent international cooperation, warning that poorly managed or rapidly advancing systems could evade human control, concentrate influence among a small number of companies or countries, and pose risks through cyberattacks, election interference, weapons development and autonomous warfare. They called for common standards covering safety testing, risk assessment, human oversight and incident reporting, while also highlighting AI’s potential benefits for creativity, discovery and individual power. Australia, Canada, Germany, Singapore, Spain, South Africa and the UAE are among the coalition’s backers, while the United States, China, Britain and France did not sign; U.S. President Donald Trump and technology adviser Michael Kratsios opposed new global governance structures or slowing AI development.