What Is Microsoft Security Copilot? Pricing, Agents & Limitations (2026)
Microsoft Security Copilot pairs OpenAI's large language models with Microsoft's proprietary threat intelligence to give security teams an AI-powered assistant for incident response, threat hunting, and vulnerability analysis. It is not a chatbot dropped into the SOC. It is an orchestration layer that connects to Defender XDR, Sentinel, Entra, Intune, and Purview, pulling real telemetry into AI-generated analysis. Here is exactly what it does, what it costs, and where it falls short.
Security Copilot at a Glance
Microsoft Security Copilot is a generative AI-powered security solution that combines OpenAI's large language models with a security-specific model trained on Microsoft's threat intelligence, which processes over 78 trillion security signals daily. It is available as a standalone portal at securitycopilot.microsoft.com and embedded directly inside Defender XDR, Sentinel, Entra, Intune, and Purview.
The core concept: you ask questions in natural language, and Security Copilot translates those questions into queries across your security stack, synthesizes the results, and presents findings in plain English. Rather than writing KQL queries by hand to hunt for indicators of compromise, you describe what you are looking for and the system generates the query, runs it, and interprets the output.
Core Capabilities
Security Copilot is not a general-purpose chatbot. Its capabilities are purpose-built for security operations center (SOC) workflows. Here is what it actually does across the products where it is embedded:
Incident Summarization
When Defender XDR flags an incident, Security Copilot generates a narrative summary: what happened, which users and devices were affected, what indicators of compromise (IOCs) were found, and the attack timeline. This replaces the manual process of correlating alerts, reading log entries, and piecing together the story. The summary is generated in natural language and exportable as a PDF report.
Script and File Analysis
Drop a suspicious script or binary into Security Copilot and it reverse-engineers the code, identifies malicious behavior, and maps techniques to the MITRE ATT&CK framework. This is particularly useful for SOC analysts who encounter obfuscated PowerShell scripts or encoded payloads and need to understand what the code does without manually deobfuscating it.
Guided Response
After analyzing an incident, Security Copilot generates step-by-step remediation instructions. It recommends specific actions: isolate this device, block this IP, revoke this user's session token, update this conditional access policy. The analyst reviews and approves each step rather than executing blindly.
Natural Language to KQL
Security analysts describe what they want to find, and Security Copilot translates the description into Kusto Query Language (KQL) queries for Sentinel and Defender. Example: "Show me all sign-ins from outside the United States in the last 48 hours where MFA was not completed" becomes a syntactically correct KQL query ready to run. The analyst can review, edit, and execute the generated query.
Device and Identity Summarization
Point Security Copilot at a device or user identity and it pulls together a complete profile: installed software, vulnerability exposure, recent sign-in activity, group memberships, conditional access policy compliance, and risk signals from Entra and Intune.
Automated Reporting and Promptbooks
Security Copilot supports automated PDF report generation for incident documentation and executive briefings. Promptbooks are saved sequences of prompts that standardize repeatable investigation workflows, so a tier-1 analyst can follow the same investigation playbook as a senior responder.
The Agent Ecosystem
Security Copilot goes beyond a single chat interface. Microsoft has built 12 autonomous agents and partnered with over 30 third-party vendors to create agents that handle specific security tasks without constant human prompting. These agents operate within the same security boundary as the rest of the Microsoft stack, inheriting your existing role-based access controls. For background on how Microsoft Copilot agents work across the broader Copilot ecosystem, see our dedicated agent guide.
Microsoft-Built Agents (12 Total)
Defender agents:
- Phishing Triage Agent – Automatically triages phishing alerts, distinguishing real threats from false positives
- Alert Triage Agent – Classifies and prioritizes security alerts across Defender XDR
- Threat Intelligence Agent – Enriches alerts with Microsoft's threat intelligence data
- Natural Language Threat Hunting Agent – Translates plain-English questions into KQL hunting queries
Entra agents:
- Conditional Access Optimization Agent – Identifies missing or misconfigured zero trust policies (204% more gaps found versus manual review, per Microsoft's vendor-reported data)
- Risky User Remediation Agent – Investigates compromised accounts and recommends remediation steps
- Access Review Agent – Reviews user access rights and flags over-provisioned permissions
- App Lifecycle Management Agent – Monitors application registrations for security risks
Intune and Purview agents:
- Policy Configuration Agent – Reviews and recommends device management policies
- Change Assessment Agent – Evaluates the impact of policy changes before deployment
- Device Removal Agent – Handles secure offboarding of devices from the environment
- Data Security Posture Management Agent (Purview) – Monitors data classification policies and sensitivity label compliance
- Alert Triage Agent (Purview) – Triages data loss prevention alerts and recommends remediation steps
Partner-Built Agents (30+)
Third-party vendors have built agents that plug into Security Copilot's plugin architecture. Notable examples:
- glueckkanja Forensic Agent – Deep forensic investigation of compromised endpoints
- adaQuest Ransomware Kill Chain Investigator – Maps ransomware attack chains and recommends containment
- Invoke Identity Workload ID Agent – Investigates workload identity and service principal abuse
NIST AI RMF Self-Assessment
Self-assess against the NIST AI Risk Management Framework
Download Free →Integrations and Plugin Architecture
Security Copilot connects to both Microsoft's native security stack and a growing list of third-party tools through a plugin-based architecture. This means the AI assistant can pull data from sources across your security environment, not only from Microsoft products.
The plugin architecture also supports custom plugins, so organizations can connect Security Copilot to internal APIs, proprietary threat feeds, or ticketing systems not covered by the built-in integrations. For guidance on securing Copilot Studio agents that connect to these data sources, see our dedicated guide.
Pricing and Licensing
Security Copilot does not use a flat per-user fee. It uses a consumption-based model built around Security Compute Units (SCUs), which are billing units that measure how much processing power your security AI tasks consume. This model is closer to how cloud compute is billed than how traditional SaaS licenses work.
| Billing Model | Rate | Details |
|---|---|---|
| Provisioned SCUs | $4/SCU/hour | Minimum 1 SCU; billed monthly based on hourly reservation |
| Overage SCUs | $6/SCU/hour | Pay-as-you-go for demand spikes above provisioned capacity |
| 1 SCU continuous | ~$2,920/month | 1 SCU running 24/7 for a full month |
| Recommended start (3 SCU) | ~$8,760/month | Microsoft's recommended starting point for most organizations |
| M365 E5 inclusion | Included | 400 SCUs/month per 1,000 user licenses (max 10,000 SCUs/month) |
| E5 overage | Throttled (hard cap) | $6/SCU pay-as-you-go coming at a future date |
| M365 E7 ($99/user/mo) | Included via E5 | E7 includes E5, so the same Security Copilot entitlement applies |
Key distinction: E5 uses a monthly pool billing model where only actual consumption is deducted from the pool, not hourly blocks. Standalone customers provision SCUs by the hour. An Azure subscription is required for standalone deployment but is NOT required for E5/E7 inclusion. E5 is currently $57/user/month, increasing to $60/user/month from July 2026. For full Microsoft Copilot pricing across all tiers, see our pricing guide.
Limitations and Considerations
Security Copilot has real constraints that security teams need to understand before deploying it in production environments.
How Security Copilot Fits the Microsoft Security Stack
Security Copilot is not a standalone security product. It is an AI layer that sits on top of Microsoft's existing security stack and amplifies the capabilities of each component. Understanding where it fits is essential for evaluating whether it adds value to your current deployment.
Defender XDR provides the detection and response engine: endpoint protection, email security, identity threat detection, and cloud app security. Security Copilot adds natural language investigation, automated incident narratives, and guided remediation on top of Defender's alert pipeline.
Sentinel is the SIEM and SOAR platform that aggregates logs from across your environment (not just Microsoft products). Security Copilot generates KQL queries for Sentinel data, summarizes complex investigations, and creates automated response playbooks.
Entra handles identity and access management. Security Copilot's conditional access optimization agent and risky user remediation agent operate within Entra to find policy gaps and investigate compromised identities.
Intune manages device compliance and configuration. Security Copilot agents review policy configurations, assess change impact, and handle device offboarding through Intune's management plane.
Purview governs data classification, sensitivity labels, and data loss prevention. Security Copilot's Purview agents monitor data security posture and triage DLP alerts.
The practical question: If you are already running Defender XDR, Sentinel, and Entra on M365 E5, Security Copilot's 400 SCU/month inclusion means you can test AI-assisted security operations at zero marginal cost. If you are not on E5, the standalone pricing starts at roughly $2,920/month for a single SCU, which is a meaningful investment for smaller security teams. Start with the E5 pool if you have it. For a broader view of Microsoft Copilot across all product lines, see our overview.
Learn More: Video Resources
Go Deeper
Resources from across Tech Jacks Solutions
Security Hub
Enterprise security frameworks and defense strategies
Threat Intelligence
This week's AI and cybersecurity threat briefings
Agent Threat Landscape
Security risks specific to autonomous AI agents
FREENIST AI RMF Self-Assessment
Self-assess against the NIST AI Risk Management Framework
CISSP Certification
The gold standard for enterprise security leadership