Microsoft just released Microsoft Agent 365 (GA), and it’s not what most people think.

It’s NOT an agent builder.
It’s a governance + security control plane for AI agents.

Key architecture layers:

Identity → Entra Agent ID
Data governance → Purview (DLP, labeling, audit)
Threat detection → Defender (pre-execution blocking via webhooks)
Observability → centralized agent registry + Agent Map

Interesting part:
Multi-cloud agent discovery (AWS Bedrock, Google Gemini Enterprise) is already in preview.

This suggests Microsoft is positioning itself as a cross-platform governance layer.

The real challenge remains:
Agent orchestration ≠ agent governance.

Full breakdown here:
[link]