METR reviews Anthropic's Risk Report, sees low misalignment risk but key disagreements
CFGeek · x · 2026-10-20
Anthropic published its second Risk Report under its Responsible Scaling Policy, including a section on "Misalignment in high-stakes settings." METR, which reviewed a nonpublic version of that section, agrees the risk of catastrophic harm from covered models' misalignment is low—but notes several significant disagreements with Anthropic's analysis.
The exchange is a notable instance of independent third-party evaluators formally engaging with a frontier lab's safety self-assessment.
More from Safety
- Joshua Saxe: Curb AI cyber misuse without blocking open weights — joshua_saxe · 2026-10-21
- What rules would you need to trust an agent with write access? A builder's draft framework — Fresh_Spread_9223 · 2026-10-21
- WhatsApp Hid Instructions in Help Pages Telling Its AI Chatbot Not to Recommend Privacy Regulators — evilsocket · 2026-10-20
- Ex-OpenAI policy chief blasts OpenAI's retreat from AI chip controls to appease Nvidia — Miles_Brundage · 2026-10-20
- Time for Guardrails on Government Too: Rethinking Who AI Policy Constrains — neil_chilson · 2026-10-20
- MCP tool-poisoning detector broken in a day—here's the context-aware fix and what it still misses — blitzcrieg11 · 2026-10-20