AI SECURITY • 331 / 397
Understand how untrusted text, tools and autonomy create new attack paths.
Output Guardrail
An output guardrail validates or filters model output before users or downstream systems receive it.
Think of it like
Think of Output Guardrail 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 Output Guardrail.
SRE lens
Reject invalid JSON or prohibited data leakage.
Remember thisAn output guardrail validates or filters model output before users or downstream systems receive it.