New tools and guidance: Announcing Zero Trust for AI

Microsoft announces its Zero Trust for AI (ZT4AI) approach, adding an AI pillar to the Zero Trust Workshop, expanding Data and Networking in the Zero Trust Assessment, and publishing a Zero Trust for AI reference architecture plus practical patterns and practices to secure AI across the model lifecycle. The release focuses on applying Zero Trust principles—verify explicitly, apply least privilege, and assume breach—to AI systems, and highlights tools, guidance, and RSAC 2026 sessions to help organizations operationalize AI security.