Microsoft announced Project Perception, a new agentic security system designed to help organizations continuously identify, evaluate, and reduce security risk.

The company also unveiled Microsoft AI-Cyber-1-Flash, Microsoft’s first cybersecurity-specialized AI model, designed for software vulnerability analysis.

As AI accelerates the speed and scale of cyberattacks, defenders are being asked to secure increasingly complex digital environments with approaches built for a different era. Security teams are often forced to piece together signals, context, and risk insights across vast amounts of data, making it harder to keep pace with emerging threats. 

To address this challenge, Microsoft is introducing a new approach that combines security context, specialized AI models, and autonomous agents to continuously identify, assess, and help remediate risk across the digital estate.

Key updates of the announcement include: 

·       Project Perception, a new agentic security system designed to help organizations continuously identify, evaluate, and reduce risk, shifting security teams from reactive response to proactive risk reduction. Perception coordinates three classes of specialized agents that work together to improve security posture over time:

o   Red agents identify potential attack paths and vulnerabilities before they can be exploited. 

o   Blue agents investigate findings, apply security context, and determine what represents meaningful risk. 

o   Green agents take corrective action and strengthen defenses across the environment.

·       Microsoft AI-Cyber-1-Flash, Microsoft’s first cybersecurity-specialized AI model, designed for software vulnerability analysis and developed as part of the company’s broader multi-model strategy that combines frontier and specialized AI models for different security tasks.

Built on Microsoft’s breadth of security signals, threat intelligence, and real-world operational experience, Project Perception is designed to help organizations move beyond reactive defense and toward a more adaptive approach to security.