← Back to Model Beat
Policy·Sep 30·all news from September 30, 2026

AI companies want to embed safety evaluators, but countries need their own

Countries currently sidelined from the global competition between the U.S. and China are seeking methods to establish independent safety standards for imported AI models. While companies like OpenAI and Anthropic are developing internal safety evaluators, these nations are exploring ways to maintain regulatory oversight and technical control as they integrate third-party technology. This push highlights a growing tension between centralized corporate safety frameworks and the desire for sovereign governance over AI deployment.

Covered by 1 source · 2 articles

Related stories

PolicyOpenAI Parts Ways With 3 Workers Over Mishandling InformationSep 29 · 14 sourcesPolicyHow we will do better for AustraliaSep 27 · 5 sourcesPolicyInsurers brace for millions in claims as AI agents spin out of controlOct 4 · 12 sourcesPolicyDisrupting a coordinated model-distillation campaignSep 30 · 3 sources