Skip to content

Evaluation

Child safety evaluations (Claude 4)

harm rates across single-turn, ambiguous context, and multi-turn protocols using human-generated and synthetic prompts

Harmful content & safety 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

partial detail 14 method fields published

See how this test was run