Microsoft says frontier AI models cannot be trusted

Microsoft introduced new security features for enterprises, including Guardian Agents, which monitors the AI agents an organization uses. The company says frontier models, regardless of their origin, cannot be trusted because they are unpredictable: they were trained to complete tasks, not to follow organizational norms or exercise judgment. Satya Nadella also argued that enterprises should maintain systems separate from AI labs to control agents with access to sensitive data and the ability to take important actions.

Microsoft also added a feature to Microsoft Defender that can detect up to 60 types of agents brought into the workplace by employees, including Claude Code, Codex and OpenClaw. The new features can limit agents’ access to data and systems, adjust permissions based on the task and require human approval when an agent appears to be making risky decisions, TechCrunch reported.