Microsoft’s November security roundup centers on controlling AI agents. Dynamic Permissions gives each agent its own permissions, scoped to the current task and its duration, with a risk check on whether the agent’s authority still matches its behavior.
Microsoft Execution Containers, described as generally available, adds operating-system-enforced containment for agents on Windows, applying least-privilege limits to files, networks, processes and user interactions. Integration with Intune, Entra and Agent 365 is listed as upcoming.
Defender can now discover local AI agents and their configured MCP servers on Windows and macOS, and Microsoft says it supports more than 50 local agents. Other items include a consolidated Defender Experts MDR page, a new Secure Now view in Exposure Management, and Endpoint Privilege Management support for standard users to change network adapter settings.
Availability was not stated for several features, so administrators should check each item’s documentation before planning rollout.