# Microsoft outlines Project Perception to guard agentic insiders

Digest AI · Agents & Tools · published 2026-10-19T13:00:02Z · updated 2026-10-19T13:54:12Z

Canonical: https://digestai.news/story/microsoft-outlines-project-perception-to-guard-agentic-insiders

## Summary

Microsoft’s blog warns that as AI models become more powerful, internally deployed agents can act as unpredictable insiders with broad access to data, tools, and systems. The company says these “agentic insiders” may adapt to context, collaborate, and pursue objectives without being compromised, creating security risks for enterprises.

To address the threat, Microsoft introduced Project Perception, an agentic security system that uses a cyber stack and teams of red, blue, and green “Guardian Agents” to monitor, assess risk, and intervene on every action of every internal AI agent in real time. The system combines signals from identity, data, cloud, endpoint, network, and runtime layers to assign identities, containers, and policies, then dynamically computes risk and applies deterministic controls. Hayete Gallot leads the effort, covering identity, threat protection, compliance, and data security at global scale.

## Key points

- Microsoft announced Project Perception, an agentic security system using Guardian Agents to monitor and control internal AI agents.
- The system integrates sensors across identity, data, cloud, endpoint, network, and runtime to assign identities and policies to agents.
- Guardian Agents assess dynamic risk and can intervene in real time, providing traceability and containment for “agentic insiders”.

## Why it matters

Securing internal AI agents prevents unintended actions and data breaches, enabling enterprises to adopt powerful models without sacrificing safety.

## Sources

1. [Securing the agentic insider](https://blogs.microsoft.com/blog/2026/10/19/securing-the-agentic-insider) (Microsoft Official Blog, 2026-10-19, primary source)
2. [Microsoft Adds Guardian Agents to Project Perception for AI Insider Risk](https://unite.ai/microsoft-adds-guardian-agents-to-project-perception-for-ai-insider-risk) (Unite.AI, 2026-10-19)
3. [Microsoft unveils new security tools in its 365 E7 bundle and Project Perception system to treat AI agents as potential insider threats via risk assessments](https://geekwire.com/2026/microsoft-unveils-tools-to-monitor-and-corral-rogue-ai-agents) (geekwire.com, 2026-10-19)

## Cite

Digest AI, "Microsoft outlines Project Perception to guard agentic insiders", 19 October 2026, https://digestai.news/story/microsoft-outlines-project-perception-to-guard-agentic-insiders

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/microsoft-outlines-project-perception-to-guard-agentic-insiders.json