Microsoft's Guardian Agents treat AI agents as insider risks to cap runaway permissions
emmanuelvivier · x · 2026-10-21
Microsoft announced Guardian Agents and new Playbooks as an extension of Project Perception, its agentic security system — built on the idea of treating frontier models like insider risks, not because they're malicious, but because any non-deterministic system needs deterministic controls.
Key points:
- Guardian Agents adapt as agent activity, risk, security context, and org policy change, helping teams understand why risk shifted, investigate evidence, and recommend or apply containment.
- Signals, context, and a unified registry are shared with Microsoft Agent 365, giving IT and security teams a consistent view of agents and their risks.
- The framework spans four ongoing duties: know, contain, protect, and adapt — covering identity, permissions, data, endpoints, networks, cloud infra, and runtime.
- People retain control over objectives and consequential decisions.
Related event: Microsoft Expands Project Perception, Treating AI Agents as Insider Risks(7 posts)→
More from coding & agent
- Redditors ask: what does it really take to keep AI agents reliable in production? — djgoel123 · 2026-10-21
- 'How Do People Code Without Claude?' — 'Easy, We Use Codex' — alexmacgregor__ · 2026-10-21
- One image, 40 nested worlds: creator shares the full workflow behind his zero-cut infinite-zoom film — tetsuoai · 2026-10-21
- Google releases free 1-hour AI engineering course, from first agent to autonomous systems — msharmas · 2026-10-21
- Feeding agents CSV/Sheets data: raw rows, SQL tools, or structured query APIs? — Extreme-Technology77 · 2026-10-21
- Developer lets an agent iterate astrophotography pipelines overnight using NXCORR comparisons — fofrAI · 2026-10-21