Date: July 27, 2026 Main source: Microsoft Link: https://blogs.microsoft.com/blog/2026/07/27/rethinking-security-for-the-age-of-ai/
Summary
Microsoft introduced Project Perception, an agentic security system designed to detect, reason about and act on threats at machine speed. The system coordinates red, blue and green team agents to discover attack paths, investigate risks and apply fixes. It uses a multi-model architecture that combines frontier models with models specialized in cybersecurity. The public preview will start on August 3, 2026.
Why it matters
Security is moving from tools that generate alerts to systems that run full detection and response cycles. Microsoft also shows that mature enterprise deployments will not depend on a single model, but on an orchestration layer that picks the right model based on cost, latency and accuracy.
Practical implications
- Any agent with access to CRM, email, files or automations needs minimum permissions, logs and the ability to roll back.
- Agencies should separate detection, analysis and execution agents, instead of giving full access to a single agent.
- A multi-model architecture can reduce costs and dependence on a single provider.