Microsoft has warned that unpredictable AI agents could become a new insider security threat to enterprises, introducing additional protections through Project Perception.
Microsoft has warned that unpredictable AI agents could become a new insider security threat to enterprises, introducing additional protections through Project Perception.
Features including Dynamic Agent Risk, Dynamic Permissions, and Guardian Agents are designed to continuously monitor agent activity, restrict access, and respond automatically to emerging threats. Microsoft emphasizes that even non-malicious AI agents can create security risks when granted access to sensitive corporate systems and data.