Microsoft Adds Dynamic Permissions, Risk Scoring to Curb Rogue AI Agents

Microsoft announced new security controls for AI agents, treating them as insiders with per-task access and dynamic risk-based permission tightening. This addresses the emerging threat of rogue AI agents acting beyond their intended scope. The move signals a shift in enterprise security posture as AI agent adoption grows.

AI disclosure: Model-written summary of the source, not a verified fact. Check the sources. How Mosaic works