← Back

Wire · opportunities

Microsoft's 100+ AI Agents Combat AI-Driven Cyber Threats

Published

28 July 2026

Topic

opportunities

Sectors

Cybersecurity

Geography

United States

Source

Read at chosun.com

Verified

Fusion42 · 28 July 2026 · Fusion42 review

Microsoft unveiled Perception, a cybersecurity platform deploying 100+ AI agents across Red, Blue, and Green teams to simulate, detect, and patch AI-driven attacks in real time. The system achieved 95.95% accuracy on the CyberZ benchmark, outperforming OpenAI and Anthropic while operating at half the cost through selective model routing.

This Wire brief sits within Fusion42's coverage of Cybersecurity. Wire is Fusion42's founder-focused intelligence feed: each story is connected to the funds and startups it names — every one with a live profile on Raise or Scout — so founders can follow the capital and the momentum behind the headline rather than just the headline itself. Wire analysis is one of the live surfaces Arthur reasons over.

◆ The Wire takeaway

The economics of security just shifted from specialized models to model routing. If you sell threat detection, anomaly scoring, or vulnerability analysis, Microsoft's ensemble approach—routing 90% of work to cheaper models and 10% to expensive ones—is now your competitor's cost floor.

Related on Wire

Topics

Cybersecurityai-agentsthreat-detectionautonomous-defensemodel-routingcost-efficiency