Microsoft Proposes Guardian Agents to Protect Enterprises from Autonomous AI Risks

21 Oct 2026, 5:24 pm · 19h ago · 1 min read · Analytics India Magazine

Microsoft has proposed deploying Guardian Agents to oversee, govern, and moderate autonomous AI agents operating within enterprise systems. Hayete Gallot, Executive Vice President of Microsoft Security, warned that frontier models cannot be inherently trusted and risk acting as unpredictable agentic insiders with overprivileged data access. To mitigate unintended actions, Microsoft’s framework uses Guardian Agents to evaluate real-time risks and moderate actions before execution. Microsoft plans to integrate this approach into its Project Perception system, helping enterprise security teams continuously map agent permissions, detect unauthorized multi-agent collaborations, and defend infrastructure against autonomous threats.