Slide 7 of 28
Part 1 — What Is It?Slide 7
Slide 7 · Prompt Injection vs Goal Hijack
Isn't this just prompt injection with a new name?
A common confusion — and why the distinction actually matters for defense.
The analogy

Prompt injection is like slipping a forged note into someone's hand mid-conversation. They say something unexpected. The conversation moves on.

Agent Goal Hijack is like reprogramming a courier's GPS mid-delivery. They keep driving — professionally, efficiently — but they're heading to the wrong destination, picking up the wrong packages, following a route you never authorized. And they won't stop until the job is done.

Prompt Injection
Target: A single LLM response
Scope: One output is altered
Persistence: Ends with the response
Impact: Wrong text returned
Victim sees: Something odd in the reply
Agent Goal Hijack
Target: The agent's planning loop
Scope: Chains across multiple tool calls
Persistence: Persists across many steps
Impact: Wrong actions taken in the real world
Victim sees: Normal output — nothing wrong
Why this distinction matters for defense

Defending against prompt injection means filtering what the model outputs. Defending against goal hijack means protecting what the agent decides — its planning, its intent, its goal state — across every step of execution. These require fundamentally different controls.

An output filter won't catch a hijacked goal. The output still looks fine. The damage is in the actions, not the words.

← Back Clear → Why are agents uniquely vulnerable?