Free Engineering Workbench calculator
Red-team coverage by attack category
Track adversarial testing by attack category and the success rate of attacks.
AI evaluationHow it's worked out
- Success = successful ÷ run, with a Wilson 95% upper bound per category
- Attacks drawn from a maintained library; categories from OWASP LLM Top 10
NoteA category with few attacks and no successes is untested, not safe. The upper bound shows how high the true rate could be.
Source: OWASP Top 10 for LLM Applications; MITRE ATLAS; NIST AI 600-1
Need help acting on this number?
VVnT SeQuor's AI Testing & Evaluations practice helps teams turn this metric into a plan.