Microsoft Adds Guardian Agents, Hayete Gallot Says Frontier Models Cannot Be Trusted
Summary
Microsoft adds Guardian Agents to Project Perception, monitoring enterprise AI agents for insider-threat behavior. Security chief Hayete Gallot says frontier models cannot be trusted because they are unpredictable. Microsoft Defender now detects up to 60 agent types, including Claude Code and Codex, and can require human approval for risky actions.
Key Points
- Project Perception includes three agent classes launched in June: red team, blue team, green team.
- Satya Nadella writes that enterprises must separate the supply of intelligence from the authority over it.
- Microsoft is an inaugural member of Nvidia's Open Agent Safety Platform, which counts more than 100 supporting tech companies.