AI SECURITY • 342 / 397
Understand how untrusted text, tools and autonomy create new attack paths.
System Prompt Leakage
System prompt leakage exposes hidden instructions or policy text intended to remain internal.
Think of it like
Think of System Prompt Leakage 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 System Prompt Leakage.
SRE lens
Leaked tool descriptions can also reveal sensitive architecture details.
Remember thisSystem prompt leakage exposes hidden instructions or policy text intended to remain internal.