RiskRubric Scoring Methodology
Released: 06/04/2026
The RiskRubric Scoring Methodology provides the technical foundation for evaluating and benchmarking the security posture of AI models, MCP servers, and AI agents. Designed to produce consistent, transparent, and reproducible risk scores, the methodology combines established risk management principles with AI-specific security testing approaches.
This document explains:
- How RiskRubric measures residual risk through structured attack-based evaluations
- The methodology used to calculate risk scores based on impact, likelihood, and attack success rates
- How NIST and OWASP frameworks are applied to create objective and repeatable assessments
- The weighting, normalization, and scoring processes used to benchmark AI systems across different risk categories
The methodology incorporates adversarial testing strategies, dynamic risk weighting, and standardized scoring thresholds to help organizations interpret evaluation results and make informed deployment decisions.
Download this Resource
Prefer to access this resource without an account? Download it now.
