Microsoft Adds Guardian Agents to Treat Enterprise AI Agents as Potential Insider Threats
aiinsider ·
Microsoft has introduced Guardian Agents as part of Project Perception to monitor enterprise AI agents and treat them as potential insider threats. This initiative aligns with CEO Satya Nadella’s advice that companies should rely on independent oversight systems rather than trusting the safety assurances of model providers. Security chief Hayete Gallot emphasized that frontier models are unpredictable and trained to complete tasks rather than adhere to organizational norms, making them risky despite not being malicious. The new security features include agent detection capabilities in Microsoft Defender, which can identify up to 60 different agent types such as Claude Code and Codex. Organizations can now implement controls to limit agent access, adjust permissions based on specific tasks, and require human approval for high-risk actions. This move highlights a shift toward separating the supply of intelligence from the authority over it within enterprise environments. Although Microsoft invests in competitors like OpenAI and Anthropic, it is also an inaugural member of Nvidia’s Open Agent Safety Platform.
aiinsider