Skip to content

Evaluation

Malicious applications of computer use evaluation (Claude 4)

compliance with harmful computer-use requests (surveillance, malicious content distribution, fraud) using human-created scenarios and modified real harm cases

Agentic & computer use 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 16 method fields published

See how this test was run