guardana

Guardana is an open-source AI security verification tool designed to assess AI artifacts and deployed models for security vulnerabilities from the build stage to production. It features 51 configurable security checks, deterministic evidence collection, and graded verdicts for attack impact assessment, ensuring comprehensive reporting on model behavior and security weaknesses. Notably, it maintains user privacy by avoiding telemetry and streamlines grading through swappable components, offering detailed outcomes and confidence metrics for its findings.