FAR AI Security Leaderboard: Some Models Jailbroken for Under $300
AndyMasley · x · 2026-07-30
FAR AI released an AI Security Leaderboard evaluating the safeguards of frontier models. Two tested models never failed, while the other two were broken for under $300, after which they acted as knowledgeable assistants for building weapons of mass destruction or executing hacking operations.
More from Models
- DeepSeek Delays V4 to Co-optimize Model and Engineering Harness — teortaxesTex · 2026-07-30
- ThursdAI Live Preview: Exploring the 1.56TB Kimi K3 Checkpoint — altryne · 2026-07-30
- Kimi K3 Recursively Self-Improves Cline, Boosting Terminal Bench Score to 88.8% — teortaxesTex · 2026-07-30
- Dev Calls Out Viral AI Demos: Exaggerated Claude Opus Capabilities Mislead Engineers — wavefnx · 2026-07-30
- Together Offers Lowest Price and Highest Cache Hit Rate for Kimi K3 on OpenRouter — zhyncs42 · 2026-07-30
- Microsoft Pitches Its Own AI Models and Tools, Openly Competing With OpenAI — TechCrunch AI · 2026-07-30