Scientists call for international AI oversight body

About 400 scientists from 34 countries, including Geoffrey Hinton and Yoshua Bengio, have called for an international body to oversee safety measures for advanced cybercapable artificial intelligence systems. The proposal, set out in an open letter dated October 13, calls for shared containment standards and a registry of significant AI security incidents. The signatories recommend isolating frontier models from external networks during development and testing, including through airgapped systems, and placing them under independent oversight. They say organizations working on cybercapable models should adopt high-assurance containment measures while governments pursue formal oversight. The proposed body would establish industrywide containment standards and maintain a record of major incidents, which the letter says should eventually be made public. The signatories argue that voluntary company commitments do not provide sufficient safeguards. The letter is a call for action; the supplied material does not establish that its recommendations have been adopted or that the proposed body exists. The appeal follows concerns about weaknesses in testing and containment.