Microsoft Introduces New Security Tools and Guardian Agents to Mitigate Risks from Unpredictable Corporate AI

Microsoft introduced new security tools on Monday to protect businesses from the growing risks associated with corporate artificial intelligence agents. The company warned that frontier AI models cannot be fully trusted because they are unpredictable and lack organizational judgment. The security rollout includes a new monitoring feature called Guardian Agents, which stems from a June initiative known as Project Perception.

Microsoft also updated its Defender product to detect up to 60 employee-introduced agents, such as Codex and Claude Code. Additional features can dynamically adjust agent permissions or require human approval for risky decisions. CEO Satya Nadella compared the risk of unpredictable AI software to an insider threat from a rogue employee.

Security business chief Hayete Gallot stated that enterprises must separate the supply of intelligence from the authority that controls it. The approach creates a market for monitoring tools independent of AI model providers. Microsoft is also participating in Nvidia's Open Agent Safety Platform, which uses hardware controls to isolate and monitor agents.