METR says Anthropic redacted most of the key evidence from its safety assessment

YashasGunderia · x · 2026-11-06

The post shares METR's redaction summary from its assessment of Anthropic:

• Under the agreement, Anthropic could redact non-public information. METR states it could share top-level risk conclusions freely, but Anthropic redacted the majority of the evidence METR found most informative, described as "removing confidential information that could accelerate other AI developers"
• Some redactions are hard to even describe without revealing sensitive information
• METR and Anthropic researchers both have significant uncertainty about how to synthesize the redacted evidence, which had multiple plausible interpretations
• Consequently the public cannot judge the evidence themselves — METR admits the report does not achieve "transparency about evidence," and readers must rely heavily on METR's judgment

This raises a substantive debate about transparency limits of third-party RSP audits at frontier labs.

Related event: METR's first inside-data assessment finds AI has roughly doubled Anthropic's R&D speed(21 posts)→

Original post →