Microsoft is aware of AI issues and is introducing a feature to protect against threats

Microsoft has announced several times what the future of Windows 11 and later versions of the operating system will look like. We're heading towards working with AI that can see and hear everything a computer user does, making the mouse and keyboard pretty much unnecessary for using a laptop or PC. However, such extensive implementation of AI has its drawbacks, of which Microsoft is aware.

Microsoft is implementing safeguards for AI needs

Copilot is being added to more Windows 11 features, like the search bar and Excel. But that's not the end, as AI agents are expected to be active in various areas of the operating system. This creates certain issues, which the company is aware of, stating directly on Microsoft's support page:

The cross-prompt injection mentioned by Microsoft is a new category of threats aimed at misleading artificial intelligence. The attacker can sneak in harmful stuff into interface elements or documents, tricking the AI into doing things it wasn't supposed to do. This could involve, for example, data theft or the installation of malicious software.

To prevent this, Microsoft has introduced Agent Workspace, an isolated environment for AI agents. The user can access it through their account, and the AI inside will have its own way of working and permissions. This way, AI will only have access to specific folders, and the path to sensitive areas of the operating system will be completely blocked.

Agent Workspace is currently in the testing phase and is optional as an experimental feature. Users can set up accounts for AI agents and create dedicated workspaces for them. The AI might need access to the six most commonly used folders: Documents, Downloads, Desktop, Music, Pictures, and Videos.

• Microsoft
• tech
• AI / Artificial Intelligence
• Windows 11