AI Security 101: Mapping the AI Attack Surface

This document explains the emerging AI attack surface—spanning training data, models, pipelines, APIs, and shadow AI—and how it introduces risks beyond traditional cloud security, including prompt injection, data leakage, and model poisoning; it cites recent industry examples and recommends practical steps such as mapping AI usage, securing training data, hardening ML infrastructure, and monitoring AI endpoints. It also positions Wiz’s AI security capabilities (discovery, context via the Security Graph, prioritization, and AI-SPM) as a horizontal approach to gain end-to-end visibility and reduce risk across the AI lifecycle, emphasizing shared responsibility and proactive governance.