Microsoft Unveils Project Perception for Agentic Cyber Defense
Microsoft has introduced Project Perception, a multi-model security system designed to combine signals, context, models, agents and enforcement tools so organizations can respond to machine-speed threats.
Why it matters: Security systems must increasingly reason across context and act continuously as AI reduces the cost and time required to launch attacks.
Microsoft has announced Project Perception, an agentic security system designed for an environment where attackers and defenders can both operate at machine speed.
The architecture combines security signals, organizational context, frontier and specialized models, an agent harness and actuators that can translate decisions into protective action across identities, endpoints, applications, data and clouds.
Microsoft says the system uses a multi-model strategy so each security task can be matched with an appropriate balance of quality and cost. Its first highlighted scenario uses MAI-Cyber-1-Flash inside a vulnerability-management team of agents. Microsoft reported a 96% score on CyberGym and nearly 50% lower cost than the prior configuration.
Project Perception is scheduled to enter public preview on August 3. The product illustrates how major security vendors are redesigning platforms around autonomous defense rather than adding isolated assistants to existing alert systems.