Microsoft ships MXC to sandbox AI agents on Windows

Microsoft has made Microsoft Execution Containers (MXC) generally available, isolating AI agents from the rest of the system. MXC lets developers and admins set permissions for file, network and clipboard access, while Microsoft Defender can block malicious prompts before they run. The tool is open source on GitHub and works across Windows, Linux, macOS and WSL.

• MXC isolates AI agents and enforces specific permissions and policies
• Agents can be blocked from files, network, clipboard and desktop access
• Microsoft Defender will detect malicious prompts before execution
• MXC is open source on GitHub and supports Windows, Linux, macOS and WSL