tech
December 3, 2025
Microsoft tries to head off the “novel security risks” of Windows 11 AI agents
Agents with read/write access to your files create big security, privacy issues.
TL;DR
- Windows 11 is integrating experimental 'agentic' AI features, including Copilot Actions, to automate background tasks.
- These AI agents are designed to act as digital collaborators, enhancing user efficiency.
- Potential risks include errors, confabulations, and novel security threats like hijacking and data exfiltration.
- Safeguards include separate user accounts for agents, user approval for data requests, and observable actions.
- Agents can access read/write permissions for common user folders and all installed apps.
- Microsoft aims to learn from past issues, like the Windows Recall rollout, by offering these features as optional and off by default.
- New 'human-centered' Copilot features include an animated character named 'Mico' and improved voice input understanding.
Continue reading
the original article