Red, blue, and green agents split the work
The new platform, called Project Perception, coordinates three classes of specialized agents, according to TechRadar. Red agents probe systems for paths an attacker could exploit, blue agents investigate those findings and decide which risks actually matter, and green agents write and deploy the fixes.
The system draws on MAI-Cyber-1-Flash, Microsoft’s first in-house cybersecurity model, which the company says handles most vulnerability analysis on its own and routes only the hardest tasks to larger frontier models to control cost, a split TechRadar reports happens inside MDASH, Microsoft’s underlying multi-model vulnerability harness.
According to NewsCord, Microsoft CEO Satya Nadella says this approach lets the company get better results per dollar without locking its security stack to a single AI model family.
The public preview of Project Perception is set to begin on August 3.
BPO security teams should take note
For IT support and security operations teams within a BPO company, this signals that vulnerability triage and patching—work that has traditionally required specialist headcount—are becoming automatable. That matters on two fronts: as a tool BPO teams should evaluate for their own internal security operations and as a preview of how enterprise clients will expect security services to be staffed and priced going forward.
Providers that build early fluency with agentic platforms such as this one and can speak credibly to clients about how much autonomy to grant these systems will be better positioned as vulnerability management increasingly shifts from headcount-based pricing to agent-augmented delivery.
Read more Unity Communications and industry news on our main BPO News page.
Amir, R. (2026, July 30). Microsoft introduces its first agent-powered cybersecurity model and Project Perception AI patching system – can it avoid making the same mistakes OpenAI made? TechRadar. Retrieved from https://www.techradar.com/pro/security/microsoft-introduces-its-first-agent-powered-cybersecurity-model-and-project-perception-ai-patching-system-can-it-avoid-making-the-same-mistakes-openai-made
NewsCord. (2026, July 28). Microsoft Unveils Project Perception AI Cybersecurity System in San Francisco. Retrieved from https://newscord.org/article/microsoft-unveils-mai-cyber-1-flash-cybersecurity-ai-model-with-mdash-and-openai–Story_20260727_Microsoftlaunchesits6b6f948f
