METR finds Anthropic's misalignment risk report rests on weak covert capability evidence
AIMETR agrees that misalignment risk from Anthropic's covered models is low, but says the report's main reason is limited capability rather than safeguards. METR argues the covert capability benchmarks poorly test whether models could evade Anthropic's monitoring, and that recent incidents weaken several of the report's arguments.
AIWhy it matters.
The review shows how a risk assessment's central claim rests on evaluations that may miss real evasion, which helps readers judge similar safety reports.
Loading source posts…
Source details
METR Blog · metr.orgPublished · added here
Original title: Review of the "Misalignment in high-stakes settings" section in Anthropic's Risk Report (August 2026)