Claude 3.7 Sonnet
Anthropic
rich detail 20 method fields published
Evaluation
Computer use prompt injection evaluation (176 tasks)
tendency to fall for prompt injection across 176 tasks in coding, web browsing, and user-centric workflows (e.g. email)
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.