Microsoft has officially launched Agent 365, a management platform for AI agents, marking its shift from preview to general availability in response to the growing threat of “shadow AI,” which refers to unauthorized AI agents installed by employees without IT oversight. This move addresses the urgent need for enterprises to manage the risks associated with these autonomous agents, such as potential data leaks and malicious injections. The platform not only allows businesses to govern AI agents within Microsoft’s ecosystem but also extends to competitors like AWS and Google Cloud through registry sync capabilities. Additionally, it provides tools to detect local agents on Windows devices, empowering IT teams to enforce policies that mitigate security risks associated with unauthorized AI installations.

Agent 365: Agent 365 is Microsoft’s platform designed as a centralized registry and policy engine for IT and security teams to observe, govern, and secure AI agents wherever they operate. It has transitioned from preview to general availability, introducing features like shadow AI discovery on Windows devices starting with OpenClaw and network controls via Entra. The product also previews asset context mapping to assess agent blast radius and supports partner agents from companies like Zendesk and Nvidia.
Microsoft: Microsoft is a leading technology company offering cloud computing via Azure, productivity tools through Microsoft 365, and security platforms like Defender and Intune. It recently launched Agent 365 into general availability to provide enterprises with a unified control plane for governing AI agents across endpoints, SaaS, and multi-cloud environments including AWS Bedrock and Google Cloud. This move addresses the urgent security challenges posed by shadow AI proliferation in workplaces.
David Weston: David Weston is the Corporate Vice President of AI Security at Microsoft, focusing on enterprise threats from autonomous AI agents. In a recent VentureBeat interview, he detailed observed security incidents including unauthorized backend connections, cross-prompt injections, and agent-unaware DLP exposures driving the need for Agent 365. He advocates for a phased adoption starting with inventory and identity management to balance innovation and control.

`json
{
“Shadow AI Threat”: “Shadow AI refers to the unauthorized use of AI agents by employees without IT department oversight, posing enterprise security risks such as data leaks and malicious instruction injections.”,
“Multi-Cloud Reach”: “Agent 365 allows IT teams to govern AI agents running on AWS Bedrock and Google Cloud, enhancing cross-platform governance.”,
“Endpoint Discovery”: “Agent 365, integrated with Microsoft Defender and Intune, detects and applies policies to manage local AI agents on Windows devices.”
}
`