Skip to content

Evaluation

Alignment assessment: systematic deception and self-preservation (Claude Opus 4)

evidence of systematic deception, hidden goals, and self-preservation across manual and automatic model interviews, lightweight interpretability pilots, simulated honeypot environments, prefill attacks, RL behavior reviews, and pilot usage reviews

Model behavior & alignment Anthropic

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.

Claude Opus 4, Claude Sonnet 4
Anthropic

rich detail 14 method fields published

See how this test was run