Microsoft introduced new security features on Monday to protect enterprises from the growing threat of AI agents, warning that frontier models cannot be trusted and act like insider threats. The centerpiece of this launch is Guardian Agents, a monitoring feature that stems from Project Perception, an initiative Microsoft introduced in June. Project Perception originally featured red, blue, and green teams of agents designed to hack, detect, and fix vulnerabilities, operating under the assumption that threats would originate outside the company. However, Microsoft leadership now views deployed AI agents as potential insider threats that require independent oversight and control.
Microsoft CEO Satya Nadella warned earlier this month that enterprises must stop blindly trusting AI model providers, even when safety assurances are given, and urged organizations to separate the supply of intelligence from the authority over it. Hayete Gallot, Microsoft's security business chief, echoed this sentiment during the Guardian Agents announcement, explaining that frontier models are unpredictable and lack organizational judgment despite having access to sensitive enterprise data and critical tools. To address these risks, Microsoft has added agent detection capabilities to Microsoft Defender to track up to 60 workplace agent types like Claude Code, Codex, and OpenClaw, alongside new permission-limiting and human-approval safeguards.
This move highlights Microsoft's shifting strategy amid its deep ties to major labs like OpenAI and Anthropic, positioning itself as an independent security provider for enterprise AI governance. By offering tools that monitor agent behavior separately from the models themselves, Microsoft is capitalizing on the growing industry consensus that autonomou