Review of the Anthropic Sabotage Risk Report: Claude Opus 4.6
Source
metr.org
Date
Why it matters
An independent critique of how a lab argues its frontier model is unlikely to sabotage, showing which claims reviewers found well supported and which were not.
