Microsoft Sets Project Perception Public Preview for August 3
The agentic security system divides discovery, investigation and remediation across red-, blue- and green-team agents while keeping human control in the workflow.
Microsoft announced Project Perception on July 27 as an agentic security system for coordinating defensive work across an organisation's digital environment. According to the official Microsoft announcement, the product is scheduled to enter public preview on August 3.
The system is designed around a closed loop rather than a single assistant. Microsoft says Project Perception brings together signals, security context, models and specialised agents, while keeping people in control of the resulting workflows. That design places the system between raw security telemetry and corrective action: its purpose is to help defenders identify meaningful risks, investigate them and strengthen protections without treating every alert as an isolated task.
How the agent roles are divided
Project Perception coordinates three classes of agents. Red-team agents look for possible paths to compromise before attackers can use them. Blue-team agents investigate activity, reason over the surrounding context and decide what represents material risk. Green-team agents take corrective actions intended to strengthen the environment.
The separation matters because it gives each part of the defensive cycle a distinct role. Discovery, investigation and remediation can inform one another, but they are not collapsed into a single opaque response. Microsoft describes the combined system as continuously learning, with human control remaining part of the operating model.
Microsoft also says the system uses a multi-model architecture rather than relying on one model for every security task. It is integrated across Microsoft Security products, linking the agent workflow to the places where organisations already manage identities, endpoints, applications, data, cloud resources and AI systems.
Why it matters
Project Perception is a useful marker for how large security platforms are adapting agent design to operational work. The important change is not simply generating incident summaries. Microsoft is proposing coordinated agents that divide offensive simulation, defensive reasoning and remediation into connected roles.
For security teams, that raises the value of governance alongside automation. A closed-loop system can act across more of the environment, so reviewers will need clear controls over which actions are suggested, which are executed and where a person must approve a change. Microsoft says the product inherits its existing security, compliance, governance and operational controls and follows its responsible-AI principles.
The August 3 public preview will be the first practical checkpoint. Readers should treat the current material as an official product announcement, not independent evidence of detection quality or operational outcomes. What matters next is how the preview exposes approvals, audit trails, model selection and measurable performance in real customer environments.
Status
Confirmed. Microsoft officially announced Project Perception and its planned public-preview date. Performance and operational impact remain to be tested during the preview.
Sources
Update note: Last reviewed 2026-07-29. We will revise this post when Microsoft publishes preview access details, evaluation evidence or deployment results.
Sources
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.