Microsoft Wants AI Everywhere. Can IT Keep It Secure?

IT is moving from securing people and applications to governing AI agents and non-human identities, while Windows itself is evolving to accommodate that shift.

Windows becomes a hybrid AI platform: Microsoft’s vision is for Windows to combine AI processing on the local device with cloud models, using Microsoft Execution Containers to isolate agents and provide stronger security controls. The big question is how far this will extend beyond Microsoft’s own AI ecosystem.

Windows 11 gets Administrator Protection: Windows 11 26H2 introduces Administrator Protection, designed to provide just-in-time elevation without leaving users with standing administrator privileges. It promises a much stronger security boundary than traditional UAC, although application compatibility could create some early headaches.

The ConnectWise compromise highlights the risk of non-human identities: A supply-chain-style incident involving a third-party application demonstrates what can happen when credentials associated with applications that have cross-tenant access are compromised. It puts renewed focus on auditing enterprise applications, permissions and credentials in Microsoft Entra ID.

AI agents become insider risks: As organisations give agents access to more systems and data, Microsoft is acknowledging a new category of “agentic insider” risk. Guardian agents and Project Perception could help supervise other agents, but that raises an obvious question: who watches the agents watching the agents?