WildGuard

Evals

Tool (not an eval — tooling/reference entry)

Moderation model for harm, jailbreaks, refusals.

Category
Guardrails & safety classifiers (adjacent)
Insurance exposure
Regulatory/bias
Publisher
Allen Institute for AI
Publisher type
Nonprofit / research org
License
Apache-2.0
Status
Stale (>12mo)
Last commit
2024-12-02
Stars
141

View source on GitHub →

Wanich Risk Lab, 2026.