Microsoft launches Guardian Agents to monitor enterprise AI agents

Project Perception, introduced in June, originally launched three classes of agents: a red team that acted like hackers, a blue team that detected attacks and assessed vulnerabilities, and a green team tasked with fixing them. That design assumed AI hackers would come from outside the company.

Microsoft's new features shift that assumption inward. Its security business chief, Hayete Gallot, wrote that frontier models cannot be trusted not because they are malicious or easily hacked, but because they are unpredictable, having been trained to achieve tasks rather than follow organizational norms or judgment. The company's answer is to place monitoring and permission controls in the hands of the cloud provider rather than the labs that supply the agents.

Microsoft is also extending the idea beyond its own products. It was an inaugural member of Nvidia's new Open Agent Safety Platform, which drew support from more than 100 other tech companies and isolates agents in secure environments with hardware-level monitoring and a kill-switch-like stop function.