Eight countries and major AI firms sign Paris AI safety pledge, including model shutdown mechanisms

The Decoder reported on November 10 that eight countries and major AI companies signed the Paris AI Safety and Security Initiative on November 9. Signatories pledged shutdown mechanisms for models they control, common evaluation procedures for highly capable models, publication of non-sensitive test results, disclosure of known risks, and support for independent audits. Germany did not sign. The declaration is a political self-commitment with no deadline or sanctions; shared standards and cooperation procedures are still to be drafted, and no binding international oversight system exists yet.

Written by AI from the coverage · updated 2 days ago