Considerations for High-Risk AI Systems
High-risk AI systems require stronger, more specialized safeguards. Given their complexity, this section of the guide is dense and begins with a simple orientation to help users navigate the structure and locate the material they need quickly.
Section Overview
-
What Is a High-Risk AI System and Why It Matters
- Defines high-risk AI systems according to how the model is used and what it can do.
- summarizes the types of harm these systems can cause.
-
AI Architecture and Vulnerabilities of High-Risk AI Systems
- Outlines major AI system architectures and describes how each one changes the attack surface.
- Identifies pressure points where most high-risk systems fail in practice.
-
How to Secure High-Risk AI Models and Systems
- Offers practical guidance for implementing security controls that reduce risk across different architectures and deployment patterns.
- Describes baseline controls for all high-risk systems and architecture-specific control modifiers.
- Provides a detailed control library that maps mitigations to threats aligned with OWASP Top 10 and prioritizes them by potential consequence.
Because these concepts build on one another, some intentional overlap will help reinforce key ideas as we move into defining what constitutes a high-risk AI system.