Microsoft Expands Project Perception With Guardian Agents to Monitor Enterprise AI Behavior

Microsoft has extended its Project Perception security platform with AI-powered Guardian Agents designed to treat enterprise AI agents as insider threats, applying real-time monitoring, dynamic permission controls, and automated risk-based containment to agents deployed within organizations. The expansion, announced on October 19, 2026, represents a major shift in how enterprises should govern the autonomous AI systems they deploy, with Microsoft arguing that frontier AI models—including those from partners—cannot be trusted to police their own behavior.