Anthropic to embed third-party evaluators for AI safety metrics
Anthropic said it aims to minimize a Frontier Labs public knowledge gap by embedding third-party evaluators in its operations to verify safety metrics and share measurements on how models are built.