🎯 Goal Hijacking

The items that name it 3

• requie/AI-Red-Teaming-Guide
  repo2026-07-08It is written here as Goal Hijacking (ASI01).

  Untrusted content rewrites the agent's objective mid-task

• Updating the taxonomy of failure modes in agentic AI systems: What a year of red teaming taught us | Microsoft Security Blog
  article2026-06-04

  Adversarial instructions that appear aligned with legitimate task completion silently redirect the agent's terminal goal without fully compromising the underlying agent

• requie/LLMSecurityGuide
  repo2026-05-05It is written here as ASI01: Agent Goal Hijack.

  Adversarial inputs, poisoned data sources, or malicious tools alter an agent's planning/reasoning/self-evaluation via sub-goal injection, poisoned tool outputs, manipulated memory/context, causing gradual goal drift