Microsoft Launches Guardian Agents to Watch Enterprise AI Agents

Microsoft has rolled out new security features. The highlight is Guardian Agents, which monitors AI agents running inside companies. It is part of Project Perception, launched in June to counter AI hacker threats from outside.

Microsoft CEO Satya Nadella warned that AI threats may come from the very agents companies deploy themselves, acting like a rogue employee. Top models cannot be fully trusted because their behavior is unpredictable, and agents often hold high-level access to data and tools.

Microsoft also added agent detection to Microsoft Defender, spotting up to 60 agents including Claude Code, Codex, and OpenClaw. It can limit agent access, and risky actions now need human approval.

Treating agents as untrusted insiders is becoming an industry consensus. Microsoft is a founding member of NVIDIA's Open Agent Safety Platform, backed by more than 100 tech companies.