Microsoft Ignite 2026: Windows adds security for AI agents
Microsoft said its Windows endpoint security platform and administrator protection are now generally available, and it detailed new tools to run AI agents more safely on Windows.

Key takeaways
- Microsoft says the Windows endpoint security platform and Windows administrator protection are now generally available for all customers.
- Windows Autopatch customers deploy updates more than 40% faster than those using on-premises management, across more than 150 million enrolled devices.
- New containment tools in Microsoft Execution Containers aim to run AI agents safely, with OpenAI and Meta building on the technology.
- Windows Server 2028 is in public preview, and Windows baseline protection is in preview in the Windows Insider Program.
In this article
Windows security tools go live
At Microsoft Ignite 2026, the company detailed how it plans to keep Windows secure and ready for AI agents at work. In a post on the Windows Experience Blog dated 17 November 2026, Pavan Davuluri, executive vice president of Windows and Devices, said two security features, the Windows endpoint security platform and Windows administrator protection, have reached general availability, so all customers can use them.
The endpoint security platform lets security firms move important protections out of kernel mode, the core of the operating system, and into user mode. Microsoft says this makes Windows more resilient, because one failure is less likely to affect the whole device. Partners building on it include Bitdefender, CrowdStrike, ESET, SentinelOne, Sophos, Trellix and TrendAI.
Administrator protection, also generally available, lets IT staff run local administrators with standard user rights by default. When a task needs admin rights, Windows Hello checks the user and grants those rights for that task only. The extra rights end when the task ends.
A third feature, Windows baseline protection, is in preview in the Windows Insider Program. It brings together protections such as secure boot, virtualization-based security and memory integrity into a single security state. That gives IT a yes-or-no check on the security of each device in a fleet.
A faster-moving threat
Microsoft also pointed to a shift in the threat landscape. The company's latest Digital Defense Report found that turning a software flaw into a weapon, once a job for skilled experts, can now be done with a single prompt to an AI tool. Microsoft says the number of new security flaws across the industry is on course to double this year.
To keep up, Microsoft is using AI itself. Microsoft says its scanning tool, MDASH, now checks Windows code at a larger scale than before, so flaws are found and fixed sooner. In October alone, Microsoft issued fixes for over 1,100 vulnerabilities, more than it fixed in the whole of 2025.
Those fixes only help once they reach devices. Microsoft says its fastest customers secure over two-thirds of their devices within seven days.
Autopatch and app updates
Windows Autopatch, Microsoft's service for rolling out updates, is central to that effort. The company says Autopatch customers roll out updates more than 40% faster than those using on-premises management. Usage has climbed more than 50% over the past year, reaching more than 150 million enrolled devices.
Autopatch now hands IT more control over which updates go out and which features they include. Microsoft says it also cuts interruptions to a single reboot each month.
The same approach is spreading to apps. Microsoft Intune Enterprise App Management can now preview automatic updates not only for apps that IT deploys, but for apps people install themselves, using the controls and deadlines IT already knows. Windows is also moving to declarative management, which keeps devices in the state that IT sets. Policies like Wi-Fi and VPN apply within minutes and resist drift, even if a device is offline.
Making Windows safe for AI agents

Much of the announcement focused on AI agents, software that can carry out tasks across files, apps and networks on a user's behalf. Microsoft says agents are the next big shift in computing and wants Windows to be the most secure place to run them.
New containment features in Microsoft Execution Containers, or MXC, are meant to limit what an agent can reach. Microsoft says OpenAI and Meta are both building on it. Windows 365 for Agents now qualifies for Copilot Credits.
Under a banner it calls Hybrid Intelligence, Microsoft wants AI to choose the best model and processing power for every task, running that work either on the PC or in the cloud. GitHub Copilot Fusion, now in preview, can send coding tasks to local models running on a developer's PC. Microsoft says local context, along with actions and models, will come to Copilot on Copilot+ PCs soon.
One estate from PC to Azure
Microsoft also described plans to link the PC and the cloud more closely. Direct Stream for Windows 365 is aimed at that connected approach. Windows Server 2028 is in public preview. It comes with Microsoft Foundry and works with new AI hardware from chip makers AMD and NVIDIA.
For IT teams, the theme is that security, updates and AI management are being brought together under one set of controls. Microsoft frames the changes as a way to adopt agents without letting cost or risk grow faster than the value they deliver. The company says Windows now runs on more than a billion devices worldwide.
Sources
- Microsoft Ignite 2026: Building Windows for the next era of workWindows Experience Blog
Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.



