![I recently wrote about why we need to treat models as insider risks, with security controls sitting outside the model itself. [Read more]](https://i0.wp.com/techcapsules.com/wp-content/uploads/2026/10/article_image-4685.webp?fit=1024%2C576&ssl=1)
linkedin.com
Oct 20, 2026
I recently wrote about why we need to treat models as insider risks, with security controls sitting outside the model itself. [Read more]
AI-summarised brief · reviewed before publication
Microsoft’s Satya Nadella announced an expansion of Project Perception, an agentic security system designed to protect AI insiders. The update adds continuous risk assessment, adaptive permission controls, and threat containment for AI agents that operate across enterprises with excessive privileges. Perception employs a cyber stack with coordinated red, blue, and green teams to monitor and adjust AI behavior in real time, aiming to align AI operations with organizational norms while maintaining rapid response to external AI-driven attacks.
💡 Why It Matters
- · By enforcing deterministic controls on non‑deterministic AI, Microsoft addresses the growing threat of internal AI misuse, ensuring that automation does not outpace governance.
- · This approach balances innovation with security, setting a new standard for enterprise AI oversight.
