UN-backed body proposed to set safety standards for frontier AI models

More than 20 world leaders have called for mandatory safety testing and common standards for frontier AI models, proposing a UN-backed international body to oversee the most advanced systems. The September 21, 2026 statement, led by Finland and Norway, urges AI companies to adopt transparent protocols and independent evaluations. It also asks governments to coordinate standards and share incident reports, with 22 leaders endorsing it, though the US, China, Britain, and France have not. The proposal, welcomed by UN Secretary-General António Guterres, is non-binding and lacks technical thresholds or enforcement details. It remains open for additional endorsements, reflecting a push for international cooperation amid concerns that AI development outpaces safety measures.