Microsoft Launches Perception Agentic Security Platform from Preview
Microsoft has transitioned its Perception agentic security platform from public preview to general availability, marking a significant advancement in automated threat response.
General Availability of Perception Agentic Security Platform
Microsoft has transitioned its Perception agentic security platform from public preview to general availability, marking a significant advancement in automated threat response. Initially introduced during the company’s Ignite 2026 conference in July, the system has been integrated into Microsoft Defender since August.
Agent Categorization
The platform operates through specialized agents categorized into red, blue, and green teams, each designed to simulate adversarial tactics, detection processes, and remediation efforts.
Blue Team Agents
The blue team agents now generally available include functions for alert triage, attack investigation, detection authoring, and threat intelligence. Microsoft reported that the Alert Triage Agent can resolve approximately 80% of alerts in a typical Defender queue, based on analysis of around 8 million alerts across its security tools.
Red Team Agents
The Recon Agent, a red team component, is also now generally available, while a Web App Pen Test Agent remains in preview.
Green Team Playbooks
Green team playbooks have been expanded to support cross-platform operations across Microsoft, AWS, and Google Cloud. These include the Posture Prioritization Agent, which is generally available, and the Risk Reduction Agent, currently in public preview. The latter can implement temporary protections for vulnerable assets when immediate patches are unavailable, effectively shielding systems until permanent fixes are deployed.
AI-Driven Threat Response and Security Frameworks
AI Risk Management
Microsoft addressed risks associated with AI agents, which it classifies as a new form of insider threat due to their broad access to data and tools. In preview, Microsoft’s Agent 365 and Entra Suite can identify 60 types of local AI agents, including Claude Code, OpenAI Codex, and GitHub Copilot. Intune policies are being deployed to restrict unsupported agents on devices.
Enhanced Detection Capabilities
Other developments include the release of new playbooks for vulnerability management, enhanced detection capabilities for AI-driven threats, and updates to cloud security frameworks.
Integration and Automation
The platform introduces integrations with security information and event management (SIEM), security orchestration, automation, and response (SOAR) tools, as well as IT service management systems, via Microsoft Cloud Partner (MCP) and application programming interfaces (APIs).
Human Oversight and Future Developments
Control and Guardrails
Microsoft emphasized that user-defined guardrails govern automated actions, with playbooks triggered by predefined incident criteria. Aleš Holeček, president and chief architect of Microsoft Security, highlighted that teams retain control over high-impact decisions while automated workflows handle routine tasks.
Continuous Refinement
Microsoft continues to refine its approach to balancing automation with human oversight, ensuring enterprises maintain control over critical security decisions while leveraging advanced threat response mechanisms.
“Teams retain control over high-impact decisions while automated workflows handle routine tasks,” said Aleš Holeček, president and chief architect of Microsoft Security.
