The useful question is what changes for users, developers or buyers, and whether the announcement stays industry context or becomes something people can actually use.

Microsoft argues that traditional security approaches are inadequate against AI-powered attacks, which operate at machine speed and scale. The company proposes a new cyber stack capable of continuously perceiving risk, reasoning across vast data, and taking action autonomously while keeping humans in control. This system aims to adapt to evolving threats and empower defenders with deeper insights and workflows. Project Perception is positioned as a response to the changing dynamics of cybersecurity in the AI era.

Project Perception introduces a closed-loop defense system using specialized agents: red team agents identify potential attack paths, blue team agents assess risk, and green team agents enforce protections. These agents collaborate to continuously improve an organization’s security posture. The system leverages Microsoft’s visibility across identities, endpoints, and cloud environments, combined with threat intelligence and operational experience. It is designed to operate 24/7, balancing effectiveness, availability, and cost at scale.

The system employs a multi-model architecture, combining frontier and specialized AI models to optimize performance and cost. For software vulnerability management, Project Perception integrates MAI-Cyber-1-Flash into MDASH, achieving 96% accuracy on CyberGym and delivering nearly 50% cost savings compared to existing configurations. Microsoft plans to expand this model’s use across additional security workflows. Project Perception enters public preview on August 3, marking a shift toward agentic security systems tailored for AI-driven threats.

Project Perception’s cyber stack consists of layers including signals, security context, models, and actuators, all working in unison. Security context provides agents with token-efficient, near real-time understanding of an organization’s assets and risks. The multi-model approach selects the best model for each task based on quality, reliability, latency, and cost. Actuators enable automated actions, reducing risk continuously. Built on Microsoft’s Responsible AI principles, the system maintains enterprise-grade security, compliance, and governance standards.