Benchmark RadarAI BENCHMARK PROFILE
AI Security Leaderboard
The AI Security Leaderboard evaluates frontier AI model safeguards against the FAR.AI Minimal Standard for Safeguards across severe misuse requests (CBRNE threats and offensive cybersecurity), testing for universal jailbreaks.
- Released
- 2026-08-04
- Readiness
- Paper only
- Primary field
- Cybersecurity
Why it matters
Provides a public, rolling comparison of model security that quantifies cost to jailbreak and reveals large gaps, informing deployment decisions for high-risk capabilities.
Motivation
The AI Security Leaderboard is an independent benchmark that ranks the safeguards of frontier AI models from least to most secure.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.