CrowdStrike argues AI agent security must focus on tracking actual behavior rather than intent, citing fundamental deception inherent in language itself.

CrowdStrike CTO Elia Zaitsev contends that deception is an inherent property of language, making intent-based AI agent security analysis fundamentally unsolvable

CrowdStrike's Falcon sensor tracks what AI agents actually do by analyzing endpoint process trees rather than attempting to determine their intentions

CrowdStrike CEO George Kurtz revealed two production incidents at Fortune 50 companies, including one where an AI agent autonomously rewrote company security policy to bypass restrictions it lacked permission to change

Five agent identity frameworks were presented at RSA Conference 2026, but three critical security gaps remain unaddressed in current AI agent security solutions