Project Perception Expands To Guard AI Insiders
Microsoft is extending Project Perception to continuously monitor enterprise AI agents as potential insider risks, using Guardian Agents, dynamic risk assessment, and controls across identity, data, endpoints, networks, cloud, and runtime. The system keeps security authority outside the model and adapts permissions as agent behavior changes.
Microsoft is treating autonomous agents less like software features and more like privileged employees—an overdue shift as agents gain access to production systems and sensitive data.
- –Guardian Agents supervise actions continuously and can recommend or apply containment at machine speed
- –Project Perception combines cross-stack signals with Microsoft Agent 365’s unified agent registry
- –Dynamic permissions reduce blast radius when an agent’s context, behavior, or risk profile changes
- –Model transparency may improve intent detection, but Microsoft correctly frames containment as the primary control
- –Developers will need explicit agent identities, least-privilege access, auditability, and runtime monitoring from day one
DISCOVERED
1h ago
2026-10-19
PUBLISHED
2h ago
2026-10-19
RELEVANCE
AUTHOR
AGTPinsights