Microsoft adds agent discovery and Dynamic Permissions
Microsoft announced new agent controls on 19 October 2026: discovery expands by 35 types to 60 supported local agents, Entra Agent ID gains identity types for local and unknown agents, and Dynamic Permissions scopes authority to an approved task. SiliconANGLE reported no availability dates or pricing at launch.
Facts read on .
Microsoft Security published its agentic insider post on 19 October 2026. It extends discovery of local agents by 35 types, to 60 supported agents including Claude Code, OpenAI Codex and GitHub Copilot, and lets them join a unified registry shared with Microsoft Agent 365.
On identity, Microsoft says it is extending Entra Agent ID with identity types for local and previously unknown agents found by security sensors, so such an agent can become a policy target. It also introduces Dynamic Permissions: organization-wide policy sets the outer boundary, authority is scoped to the task and time-bound, and higher-impact actions can require verified human approval. Microsoft describes the Microsoft Execution Container as a sandbox that keeps controls outside the agent's reach.
SiliconANGLE reported that Microsoft gave no availability dates or pricing for the new controls, and that Conditional Access now extends to agents so rising risk can hold higher-impact actions for approval. The capability statements are Microsoft's own.
Sources
- Agents are your newest insiders. Security must move at their speed. (microsoft.com), read
- Microsoft targets the agentic insider with new controls for rogue AI agents (siliconangle.com), read