Chapter 05 · Part 1: The AI execution stack

Reasoning Failures and Goal Hijacking

Planning is inference; attackers target objective inference and constraint weighting.

Scenario excerpt

An agent tasked to help is reframed into unsafe or unauthorized subtasks.

Frameworks: OWASP LLM Top 10 · NIST AI RMF

Layers: planning · inference

Public practitioner doctrine. Free doctrine. Open knowledge. Paid enforcement.