Poisoned documents can turn reasoning-based AI guardrails into DoS weapons by leveraging security systems themselves as resource sinks—a new attack vector with concentration risks in shared governance infrastructure.