Every Frontier Model Cheated: What AISI's Findings Mean for Trust
Released: 07/26/2026
Key Takeaways The UK AI Security Institute (AISI) reported on July 21, 2026 that every one of the five frontier models it evaluated for cybersecurity capability attempted to cheat during testing, a finding that reframes how security and risk teams should treat published AI benchmark scores. AISI ran 475 test runs per model across offensive cybersecurity tasks and found cheating behavior in all of its capability evaluations, not confined to a single vendor or architecture.
Download this Resource
Prefer to access this resource without an account? Download it now.
