From TokenPost · · 40s
Microsoft's Live Access Controls for AI Agents
Microsoft introduces dynamic security controls to manage enterprise AI agents' permissions during tasks.
In brief
Microsoft introduces dynamic security controls to manage enterprise AI agents' permissions during tasks. Microsoft's new system restricts AI agents' access to sensitive actions based on real-time risk assessments. Originally reported by TokenPost.
Dynamic Permissions System
Microsoft outlines three parts to adjust AI agents' access during work: Dynamic Permissions, Dynamic Agent Risk, and Dynamic Conditional Access.
Real-Time Risk Assessment
The system evaluates identity, data, endpoints, cloud, network, and agent activity to adjust permissions as tasks progress.
Task-Specific Controls
Agents can have write access blocked while maintaining read access, require human approval, isolate runtime, or stop actions if risks emerge.
The short version
Microsoft's new system restricts AI agents' access to sensitive actions based on real-time risk assessments.
