What happened
Microsoft announced at its Ignite 2026 conference on Tuesday that Perception, a product that orchestrates groups of AI agents to carry out security work, is now generally available, SecurityWeek reported. Playbooks steer the agents, which are organized into three teams. Red agents act as attackers searching for weak spots, blue agents cover detection, triage and investigations, and green agents remediate issues.
Four blue team agents have reached general availability. They triage alerts, investigate attacks, write detections and handle threat intelligence. The red team Recon Agent is also generally available, while a Web App Pen Test Agent is now in preview. Microsoft said its Alert Triage Agent can process about 80% of the alerts in a typical Defender queue, a figure the company based on reviewing roughly 8 million alerts from its security products.
Two new green team playbooks span Microsoft, AWS and Google Cloud environments. Their Posture Prioritization agent is generally available, and the Risk Reduction agent is in public preview. When no fix exists yet, that agent can suggest vulnerability shielding, a preview feature that protects exposed assets until a patch arrives. MCP and API links to SOAR, SIEM and IT service management tools are in preview.
Context
Microsoft first revealed the system in July as Project Perception. It has been a Defender public preview since August. The company also treats AI agents themselves as a security risk. It views them as a new type of insider with wide access to data and tools. Agent 365 and Entra Suite previews can spot 60 local AI agent types, such as Claude Code, OpenAI Codex and GitHub Copilot. Intune policies can confine supported agents to the device.
Le briefing
La semaine tech, en un seul e-mail.
Les dépêches qui méritent votre temps, chaque semaine. Désinscription à tout moment.
