Seoul — Representatives from eighteen countries, including most of the world's largest AI-developing economies, signed a new AI Safety and Transparency Accord on Wednesday, establishing shared standards for pre-deployment testing of frontier AI models and mandatory incident reporting for major safety failures.
The accord creates an international panel of technical experts, modelled loosely on the Intergovernmental Panel on Climate Change, tasked with publishing an annual assessment of frontier AI capabilities and risks. Signatory nations committed to aligning national regulations with the panel's recommendations within three years.
Civil society groups gave the accord a cautious welcome, noting it lacks binding enforcement mechanisms and relies on voluntary compliance from private companies. Industry representatives, meanwhile, said the shared standards would reduce the burden of navigating fragmented national regulations across major markets.