Developing trustworthy AI: 9 key threat categories

This piece introduces the Cloud Security Alliance’s AI Controls Matrix (AICM), a comprehensive control framework mapped to leading standards (NIST, BSI, ISO) and organized across 18 domains to address nine key AI threat categories, including model manipulation, data poisoning, sensitive data disclosure, and supply chain risks. It emphasizes shared responsibility across cloud, model, and application providers; the addition of AI-specific control domains (e.g., model security); and the importance of practices like model artifact scanning and machine learning bills of materials (ML-BOMs) to enhance AI supply chain visibility and compliance.