Microsoft has unveiled new AI features in Windows 11, including ‘agentic’ AI capabilities designed to execute tasks in the background, aiming to enhance user productivity. These features, exemplified by Copilot Actions, are intended to assist users in organizing files, scheduling meetings, and sending emails seamlessly. However, the integration of these AI agents raises significant security and privacy implications.
Termed as ‘experimental agentic features,’ these AI agents operate with read/write access to user files, potentially exposing systems to novel security risks. Microsoft acknowledges the inherent challenges of ensuring the agents’ accuracy and preventing unauthorized access that could compromise user data.
To mitigate these risks, Microsoft has implemented safeguards such as assigning AI agents separate user accounts on PCs to restrict their system permissions. Currently available only in early test builds of Windows 11 and disabled by default, these features illustrate Microsoft’s approach to balancing innovation with security in its operating system.
Source: Ars Technica