Evasion Attack (AI)

An evasion attack crafts inputs that cause an AI agent's safety or policy filters to fail to detect policy violations, allowing harmful content, prompt injection, or unauthorised instructions to pass through governance controls.

Evasion attacks test the robustness of agent guardrails and are a core component of AI red-teaming exercises.

Closely related concepts include Jailbreak (AI Agent), Capability Control (AI), Policy Enforcement Point, Agent Context Isolation. Each is defined in the Prefactor glossary.