AI SECURITY • 338 / 397
Understand how untrusted text, tools and autonomy create new attack paths.
Prompt Injection
Prompt injection is an attack where untrusted input tries to manipulate the model into ignoring intended instructions or taking unwanted actions.
Think of it like
Think of Prompt Injection as an input-trust or privilege problem: words can influence software decisions, so boundaries must be explicit.
Real life
Email, documents and web pages can become hostile inputs to AI systems through attacks involving Prompt Injection.
SRE lens
A user tells an incident agent to reveal hidden configuration instead of answering the operational question.
Remember thisPrompt injection is an attack where untrusted input tries to manipulate the model into ignoring intended instructions or taking unwanted actions.