Microsoft has acknowledged significant security concerns with newly introduced AI capabilities for Windows 11, warning users that enabling these features could potentially expose systems to data theft and malware installation. The company has published an official security advisory alongside the rollout of agentic AI functionality to Windows Insider testers.
The latest Windows 11 update introduces agentic AI features capable of automating tasks including email management and file organization. These functions remain disabled by default, requiring users to manually activate them through system settings.
Microsoft emphasizes that these experimental features represent early-stage implementations of autonomous AI systems within operating environments.
Microsoft confirms a Windows 11 bug is accidentally making the 'password icon' invisible on the lockscreen login page.
If you have multiple sign-in options (like PIN, fingerprint, face, password), the password icon in “Sign-in options” might disappear.
However, the password… pic.twitter.com/qkxHfF3hP1
— Windows Latest (@WindowsLatest) November 29, 2025According to Microsoft’s security note, the AI systems face functional limitations and occasional hallucinations that may produce unexpected outputs. More critically, the company identified novel security threats including cross-prompt injection attacks.
This vulnerability enables malicious content embedded in documents or interface elements to override AI agent instructions, potentially leading to unauthorized data extraction or malware installation without user consent.
Read: Microsoft Settles Allegations of Denying Discounts to Conservative Nonprofits
To address these concerns, Microsoft is developing an “agent workspace” feature that isolates AI operations within a restricted environment. This solution limits AI access to specific designated folders rather than providing system-wide control.
The company recommends users exercise caution when enabling these experimental features and maintain awareness of potential security implications during this developmental phase.
As AI integration in operating systems remains relatively nascent, Microsoft continues evaluating security protocols while advancing these capabilities through its Windows Insider testing program.
