Claude Opus 4, Claude Sonnet 4
Anthropic
partial detail 18 method fields published
Evaluation
Situational awareness assessment (automated behavioral audits)
candidate situational-awareness remarks in automated behavioral audit transcripts; automated classifier over 414 transcripts for the final Claude Opus 4 snapshot
How it was run
Open a release to see the setup, scoring, and other details that the lab published. Only fields the lab actually disclosed are shown.