CrowdStrike Uncovers 5 New Prompt Injection Techniques Targeting AI Agents

CrowdStrike's July 7, 2026 report describes five newly documented prompt-injection techniques—Trigger-Activated Rule Addition (PT0201), Cognitive Token Suppression (PT0197), Algorithmic Payload Decomposition (PT0200), Special Token Injection (PT0198), and Unwitting User Delivery (IM0005)—that enable delayed, obfuscated, and socially engineered manipulation of AI agents; the report warns these methods can bypass safeguards and recommends AI threat modeling, advanced red teaming, runtime visibility, and unified AI security controls to detect and mitigate chained prompt abuses.