Summary
On July 27, 2026, Microsoft unveiled Project Perception, an agentic security system that coordinates red, blue, and green AI agents in a continuous loop to find vulnerabilities, triage which risks are meaningful, and write and deploy patches. It is powered by MAI-Cyber-1-Flash, Microsoft's first in-house cybersecurity model, a code-tuned derivative of MAI-Thinking-1 that scores 96% on CyberGym at roughly half the cost of leading models. The system enters public preview on August 3.
What changed
Microsoft introduced Project Perception (red/blue/green security agents) and MAI-Cyber-1-Flash, its first cybersecurity-specific model, announced July 27, 2026 with public preview beginning August 3.
Why it matters
Microsoft is operationalizing autonomous security operations, closing the loop from vulnerability discovery to remediation with specialized agents rather than a single assistant. A purpose-built, in-house cyber model at half the cost signals that security is becoming a distinct, cost-optimized model category, and pressures security vendors and hyperscalers to ship agentic detection-and-response rather than copilots that only advise.
Evidence excerpt
Project Perception is an agentic security system that coordinates three classes of AI agents in a continuous loop: red team agents that find vulnerabilities ... blue team agents that investigate ... and green team agents that fix defences ... Its MAI-Cyber-1-Flash model scores 96% on CyberGym (+12 over Mythos) at 50% lower cost ... enters public preview on August 3.