Granular sub-category controls for Prompt Injection Detection, allowing independent action and severity definitions across six attack categories.
| Where Can I Use This? | What Do I Need? |
|---|
- Prisma AI Runtime Security
|
- Prisma AIRS AI Runtime Security license
- AI Runtime Security deployed (network intercept or API
intercept)
|
Overview
The Prompt Injection Multi-Category Configurability introduces granular
control over how different types of prompt injection attacks are handled within
AI Security Profiles. Instead of applying a uniform policy across all prompt
injection vectors, administrators can tailor security actions (Allow or Block)
and assign specific alert severities (Low, Medium, High, Critical) across six
distinct attack sub-categories.
When enabling Prompt Injection Detection within an AI Security Profile,
you can select between two configuration modes:
- Standard: Applies a uniform action (e.g., Block) across all prompt
injection attack sub-categories.
- Custom: Enables category-level granularity, allowing administrators
to define discrete actions and severity ratings for each
sub-category.
Configure Prompt Injection Multi-Category
Follow these steps to configure granular actions and alert severities for
individual prompt injection attack categories in an AI Security Profile:
