Microsoft is in the process of developing Windows 11, with plans to transition it to an 'aggressive OS' in the future. However, the company has issued a warning about the potential risks associated with this new feature.
Before rolling out experimental features to Windows Insiders, Microsoft recommends that Windows 11 users only enable these features if they understand the security implications.

According to BGR , Microsoft confirmed that AI (artificial intelligence) agent components will be disabled by default because this feature can cause cross-prompt injection (XPIA) attack risks.
This attack occurs when AI applications are granted access to user files, which can lead to data leaks or malware installation. Agentic accounts, granted when AI agent features are enabled, have limited access to the user's profile folder, but still have the potential to cause unintended consequences.
Called Agent Workspace, the experimental AI agent components are now available in developer preview and will soon be rolled out to more users.
While no apps support this feature yet, Copilot will soon have access to Agent Workspace. However, this also raises privacy concerns, as Copilot can see the user's entire screen.

Agent Workspace components are considered dangerous by Microsoft and are disabled by default.
Microsoft emphasizes that security is not a temporary feature, but an ongoing commitment that will be refined over time. Agent Workspace is designed to allow AI applications to access files in the background, while users continue to use their devices.
However, users can still manage access and stop the agent when necessary. Currently, many people still express concerns and disagreements about this new feature on online forums.
Still, Microsoft believes that as more users get access to experimental AI agent features, they'll have more information about how they work and how secure they are.
Source: https://khoahocdoisong.vn/microsoft-phat-canh-bao-vi-ai-tren-windows-11-qua-ba-dao-post2149073216.html










Comment (0)