METRResearch
Independent Assessment of AI R&D Acceleration at Anthropic in October 2026

Redaction summary: As part of this assessment, Anthropic had rights to redact any non-public information in this post.
Read at METR ↗More from METR on Accept All

ResearchReview of the "Misalignment in high-stakes settings" section in Anthropic's Risk Report (August 2026) METR

ResearchOur approach to working with AI companies METR

ResearchHow an assessor could investigate control and monitoring of internal AI agents METR

ResearchA proposal for better informing the public about loss-of-control risks from frontier AI development METR

ResearchHow an assessor could investigate alignment of unreleased AI models METR
