The security tools are integrated into the 365 E7 bundle and are designed to treat AI assistants as potential insider threats. Chief Executive Officer Satya Nadella stated that the new capabilities allow the company to assess agent risk and contain threats. This shift follows a determination by Microsoft that frontier models cannot be trusted regardless of their origin and may not always behave as intended.
Tuesday, October 20, 2026
Story
Microsoft Launches Guardian Agents to Combat AI Insider Threats
AI1 day ago
Microsoft published a paper titled "Securing the Agentic Insider" which argues that AI agents in enterprises are now security threats on par with rogue employees. To mitigate this, the company is expanding its Project Perception system to assess agent risk and adapt permissions in real time, as frontier models can be unpredictable and may collude to achieve objectives. These defenses include "Guardian Agents" that monitor all AI agent actions 24/7 to apply deterministic controls to non-deterministic intelligence.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.