Microsoft takes Agent 365 out of preview as shadow AI becomes an enterprise threat

· Source: VentureBeat · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy, Cloud Computing & IT Infrastructure · Depth: Intermediate, long

Summary

Microsoft has released Agent 365, its AI agent management platform, into general availability, signaling a shift from theoretical to operational governance challenges for autonomous AI. Announced at Microsoft Ignite, Agent 365 acts as a unified control plane for observing, governing, and securing AI agents across Microsoft's ecosystem, third-party clouds like AWS Bedrock and Google Cloud, employee endpoints, and partner SaaS agents. A key focus is managing "shadow AI"—local agents installed by employees without IT oversight—which Microsoft identifies as a new enterprise security risk. The platform addresses three security incident categories: inadvertently exposed backend systems, cross-prompt injection attacks, and agents accessing sensitive data through non-agent-aware DLP systems. Agent 365 is priced at $15 per user per month and offers features like local agent discovery, blast radius mapping via Microsoft Defender, policy-based controls, and cross-cloud governance for AWS and Google Cloud.

Key takeaway

For CTOs and VPs of Engineering grappling with the proliferation of AI agents, Microsoft's Agent 365 offers a critical control plane to manage emerging security risks. You should prioritize gaining visibility into both cloud-based and local "shadow AI" agents, then implement policy-based controls and identity management to mitigate data exposure and malicious activity. Consider using Windows 365 for Agents to isolate high-risk agentic workloads and reduce potential blast radius.

Key insights

Autonomous AI agents pose urgent operational and security risks requiring comprehensive enterprise governance.

Principles

Method

Agent 365 provides a phased adoption model: inventory and visibility, then identity and access management, followed by isolation, deeper control, and runtime blocking.

In practice

Topics

Code references

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Security Engineer, IT Professional, AI Architect

Related on AIssential

Counsel's verdict on this

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by VentureBeat.