Back home

Microsoft Perception

x+web 0.92
Visit source

Intro

Microsoft Perception is the security product name Microsoft declared generally available at Ignite on 17 November 2026. The Tech Community post What's new with Microsoft Perception at Ignite 2026 is explicit: the idea introduced as Project Perception in July becomes Microsoft Perception at Ignite and goes GA. The same-day Security blog, Security for the agentic era, powered by Microsoft Perception, frames it as a new cyber stack for a world where attackers also use AI and the time from disclosure to exploit is shrinking from months to minutes.

Microsoft says this is not a chatbot bolted onto old tools. It is a stack that connects signals, intelligence, and orchestration: specialized models and agent teams share work across investigation, protection, and remediation, while security staff set priorities and approve critical actions. Red agents find and validate paths to compromise, blue agents triage alerts, investigate attacks, and turn threat intelligence into detections, and green agents identify the exposures that matter most and coordinate remediation.

This entry records the GA product name Microsoft Perception. It is not the same name as everyday Microsoft 365 security alerting, or as Copilot UI changes announced at the same event.

Advertisement

Explain

Perception is presented as defense built for an agentic era. Attackers can find and exploit weaknesses in parallel and often automatically; defense that still works “one alert, one investigation, one response” falls behind. Perception’s orchestration harness assigns work to red, blue, and green agents, and playbooks organize those agents around a specific security outcome. Teams can also describe what they need in chat and follow the work in an integrated session. Official copy stresses that critical actions still need human approval.

Project Perception in July was the vision and project name; the 17 November announcement is the name becoming a generally available product, with expanded agent capabilities. Public materials also tie it to an “agentic SOC” and a continuous discover–investigate–remediate loop. Which data sources are included, how it sits next to other Microsoft Security products, and which tenants can turn it on should be taken from official docs and the Ignite Book of News at the time.

Microsoft Perception is not a new general-purpose foundation model. It is a security product platform. It should also not be merged with same-week names such as Copilot Home or the MAI model family.

Usage

Start with What's new with Microsoft Perception at Ignite 2026 and the Security blog, then use Microsoft’s security documentation for provisioning, permissions, playbooks, and approval flows. The intended use is to pick a security outcome (a class of exposure or an alert investigation), start from a playbook or a natural-language request, let red/blue/green agents divide the work, and keep a person in the loop for critical actions.

Evaluate which signal sources it can actually see, whether agent recommendations are auditable, and whether policy can block actions that must never run automatically. Cloud regions, licensing, and anything still in preview follow the admin experience at the time you enable it.

If you only want to ask a security question in Copilot, or work an existing Defender alert queue, that is not this new product name.