04-infosec-architecture-track.md

GitHub Copilot InfoSec & Architecture Track

Document 04 | 2 Hours

Sri Bolisetty | December 2025

---

Track Overview

Attribute	Value
Duration	2 hours
Target Audience	CISO, Security Architects, Compliance Officers, Risk Managers
Prerequisites	None (technical security background assumed)
Format	Security review workshop
Core Outcome	Security policy framework and risk assessment

---

Critical Context

This track must be completed BEFORE administrators configure Copilot. Security defines the policies; administrators implement them.

flowchart LR
    subgraph SECURITY["Security (This Track)"]
        S1["Risk Assessment"]
        S2["Policy Framework"]
        S3["Compliance Mapping"]
    end
    
    subgraph ADMIN["Admin Track (After)"]
        A1["Implement policies"]
        A2["Configure controls"]
        A3["Monitor compliance"]
    end
    
    SECURITY -->|"Approved policies"| ADMIN

Loading

---

Learning Objectives

By the end of this track, security professionals will be able to:

1. Assess GitHub Copilot's risk profile against enterprise security requirements
2. Map Copilot controls to compliance frameworks (SOC 2, ISO 27001, NIST)
3. Define an organization-specific policy framework
4. Establish ongoing security review processes for AI tool releases

---

Module Structure

flowchart TB
    subgraph SEC["🔒 INFOSEC ARCHITECTURE TRACK - 2 HOURS"]
        direction TB
        M1["Module 4.1: Threat Model & Risk Assessment<br/>30 min"]
        M2["Module 4.2: Security Controls Deep Dive<br/>40 min"]
        M3["Module 4.3: Compliance Framework Mapping<br/>30 min"]
        M4["Module 4.4: Policy Framework Design<br/>20 min"]
        M1 --> M2 --> M3 --> M4
    end

Loading

---

Module 4.1: Threat Model & Risk Assessment (30 min)

4.1.1 — Understanding the Attack Surface (10 min)

Copilot Data Flow Architecture

flowchart TB
    subgraph DEV["Developer Environment"]
        IDE["IDE (VS Code/JetBrains)"]
        CODE["Local Code"]
    end
    
    subgraph GITHUB["GitHub Infrastructure (Azure)"]
        PROXY["Proxy Service<br/>(Toxicity filter, content check)"]
        LLM["Foundation LLM<br/>(OpenAI/Internal)"]
    end
    
    IDE -->|"Context extraction"| PROMPT["Prompt Construction"]
    PROMPT -->|"HTTPS/TLS"| PROXY
    PROXY -->|"Filtered prompt"| LLM
    LLM -->|"Suggestion"| PROXY
    PROXY -->|"Post-processing"| IDE
    
    CODE -->|"Read by extension"| IDE

Loading

Data Elements Transmitted

Data Element	Transmitted?	Retention (Business/Enterprise)
Current file content	Yes	Not retained
Open tab content	Yes	Not retained
Cursor position	Yes	Not retained
File path/name	Yes	Not retained
User engagement data	Yes	2 years
Suggestions accepted/rejected	Yes	Not retained

Data NOT Transmitted

• Closed files
• Files in .copilotignore
• Content exclusion paths
• Local git history
• Credentials (in theory—validated by filters)

---

4.1.2 — Threat Modeling (10 min)

Primary Threat Vectors

flowchart TB
    subgraph THREATS["Threat Categories"]
        T1["Data Exfiltration<br/>Code/secrets leave org"]
        T2["Supply Chain<br/>Malicious suggestions injected"]
        T3["IP Leakage<br/>Proprietary code exposed"]
        T4["Compliance Violation<br/>Regulated data processed externally"]
        T5["Credential Exposure<br/>Secrets in code context"]
    end

Loading

Risk Assessment Matrix

Threat	Likelihood	Impact	Residual Risk (w/ Controls)
Code transmitted to third party	Certain	Medium	Low (in-memory only, no retention)
Secrets in suggestions	Possible	High	Medium (filters, not guaranteed)
Public code in suggestions	Possible	Medium	Low (public code filter)
Malicious code suggestions	Unlikely	High	Medium (requires human acceptance)
Model training on your code	None (Business/Enterprise)	N/A	None

Mitigating Controls Summary

Threat	Primary Control	Secondary Control
Data exfiltration	No prompt retention	Content exclusion
IP leakage	Public code filter	Code review process
Credential exposure	Content exclusion	Secret scanning
Compliance violation	Data residency (Azure)	Audit logging

---

4.1.3 — Risk Acceptance Decision Framework (10 min)

Questions for Risk Committee

1. Is transmitting code context to GitHub/Azure acceptable?

   • Code is processed in-memory, not stored
   • GitHub is SOC 2/ISO 27001 certified
   • Business/Enterprise: No training on your code

2. Can we accept AI-generated code in production?

   • All suggestions require human acceptance
   • Existing code review processes apply
   • Static analysis/security scanning unchanged

3. What code categories require exclusion?

   • Classified/regulated code
   • Third-party code under NDA
   • Infrastructure secrets

Risk Acceptance Template

## GitHub Copilot Risk Acceptance

**Date:** [Date]
**Approved By:** [CISO/Risk Committee]

### Accepted Risks
1. Code context transmitted to GitHub infrastructure for processing
   - Mitigation: No retention, SOC 2 certified, content exclusion
   
2. AI-generated code suggestions may contain bugs
   - Mitigation: Human review, existing SDLC controls

### Conditions
- Content exclusion enabled for: [list]
- Public code filter: Enabled
- Agent Mode: [Enabled/Disabled pending review]
- Quarterly review of security posture

### Signature: _______________

---

Module 4.2: Security Controls Deep Dive (40 min)

4.2.1 — GitHub's Security Architecture (10 min)

Trust Center Overview

Location: copilot.github.trust.page

Certifications Held (December 2025)

Certification	Scope	Verification
SOC 1 Type 2	Financial controls	Annual audit
SOC 2 Type 2	Security, Availability, Confidentiality	Annual audit
SOC 3	General use report	Public
ISO/IEC 27001:2022	ISMS certification	Annual audit
CSA STAR Level 2	Cloud security (CCM)	Third-party attestation
TISAX	Automotive sector	Relevant for automotive

Responsible AI Framework

GitHub follows Microsoft's Responsible AI Standard:

• Accountability
• Transparency
• Fairness
• Reliability & Safety
• Privacy & Security
• Inclusiveness

Aligned with NIST AI Risk Management Framework (AI RMF).

---

4.2.2 — Enterprise Security Controls (15 min)

Control Categories

flowchart TB
    subgraph CONTROLS["Enterprise Security Controls"]
        direction TB
        A["Policy Controls<br/>What features are allowed"]
        B["Access Controls<br/>Who can use Copilot"]
        C["Data Controls<br/>What code is processed"]
        D["Audit Controls<br/>What is logged"]
    end

Loading

Policy Controls Detail

Control	Options	Recommendation
Public code filter	Enabled / Disabled	Enable (reduces IP risk)
Agent Mode	Enabled / Disabled	Start disabled, enable after review
Coding Agent	Enabled / Disabled	Requires careful review
Third-party LLMs	Allow / Block	Block until vendor review
Copilot Extensions	Allow / Block	Block initially
Preview features	Enabled / Disabled	Disable for production

Access Controls Detail

Control	Implementation
SSO/SAML	Enterprise: Required
SCIM provisioning	Automate seat assignment
IdP group mapping	Team-based assignment
IP allowlisting	Enterprise network only (optional)

Data Controls Detail

Control	Scope	Syntax Example
Content exclusion	Org/Repo level	**/.env, **/secrets/**
Repository restrictions	Org level	Exclude specific repos
File pattern exclusion	Global	**/*.pem, **/credentials.*

⚠️ Critical Limitation

Agent Mode and Coding Agent do NOT respect content exclusions. If content exclusion is critical, consider disabling these features.

Audit Controls Detail

Log Type	Retention	Export Options
Audit log (Copilot events)	180 days	CSV, JSON, GraphQL, REST
Git events	7 days	Same
Streaming to SIEM	Real-time	Splunk, Azure Event Hub

---

4.2.3 — Data Retention Policies (8 min)

Retention by Access Point

Access Point	Prompts/Suggestions	User Engagement	Feedback
IDE (Business/Enterprise)	Not retained	2 years	As needed
GitHub.com Chat	28 days	2 years	As needed
Coding Agent	Session lifetime	2 years	As needed

Data Residency

• Primary processing: GitHub-owned Azure tenants
• Traffic routing: Based on capacity, not geography
• No guarantee of regional data residency

For Regulated Industries

If data residency is mandated:

1. Document the Azure processing model
2. Assess if "in-memory only" satisfies requirements
3. Consider on-premises alternatives (e.g., JetBrains AI Enterprise) if required

---

4.2.4 — Audit Log Analysis (7 min)

Copilot-Specific Audit Events

Event	Trigger	Use Case
copilot.seat_assignment_created	License assigned	Track provisioning
copilot.seat_assignment_deleted	License removed	Track deprovisioning
copilot.cfb_seat_cancelled	Business seat cancelled	Billing audit
copilot.content_exclusion_changed	Exclusion rules modified	Policy change tracking

Audit Log Query Example

actor:username action:copilot created:>2025-01-01

SIEM Integration Options

SIEM	Integration Method
Splunk	GitHub App for Splunk + Audit Log Streaming
Microsoft Sentinel	Azure integration + audit log connector
DataDog	GitHub integration + webhook
Elastic	REST API + webhook

---

Module 4.3: Compliance Framework Mapping (30 min)

4.3.1 — SOC 2 Mapping (10 min)

Trust Services Criteria Alignment

TSC	GitHub Copilot Control	Evidence
CC6.1 (Logical access)	SSO/SAML, SCIM, IdP groups	Audit logs
CC6.6 (System boundaries)	Content exclusion, network config	Policy settings
CC6.7 (Data transmission)	TLS encryption, Azure network	Trust Center
CC7.2 (Monitoring)	Audit logging, SIEM streaming	Audit exports
CC8.1 (Change management)	Policy cascade, approval workflows	Enterprise settings
PI1.1 (Privacy notice)	Data handling documentation	Trust Center

SOC 2 Audit Preparation

For your organization's SOC 2:

1. Document Copilot as a subservice (SaaS)
2. Reference GitHub's SOC 2 Type 2 report
3. Document your configuration choices
4. Maintain audit log exports

---

4.3.2 — ISO 27001 Mapping (10 min)

Annex A Control Alignment

Control	Title	Copilot Implementation
A.5.15	Access control	SSO, SCIM, policy cascade
A.5.23	Cloud service security	GitHub SOC 2/ISO 27001 certs
A.8.3	Information access	Content exclusion, repo restrictions
A.8.10	Information deletion	Retention policies (no retention)
A.8.12	Data leakage prevention	Content exclusion, code filters
A.8.15	Logging	Audit logs, SIEM streaming
A.8.16	Monitoring	Metrics dashboard, alerts

ISO 27001 Documentation Requirements

1. Risk assessment including AI tools (1.3 in this track)
2. Statement of Applicability update for AI processing
3. Third-party supplier assessment for GitHub
4. Incident response procedures for AI tool issues

---

4.3.3 — NIST Frameworks (10 min)

NIST CSF 2.0 Mapping

Function	Category	Copilot Control
IDENTIFY	Asset Management	Seat inventory, license tracking
PROTECT	Access Control	SSO, policy enforcement
PROTECT	Data Security	Content exclusion, encryption
DETECT	Continuous Monitoring	Metrics dashboard, audit logs
RESPOND	Analysis	Audit log investigation
RECOVER	Recovery Planning	Seat management, policy restore

NIST AI RMF Alignment

GitHub claims alignment with NIST AI RMF through Microsoft RAI Standard:

AI RMF Function	GitHub Implementation
GOVERN	Responsible AI Impact Assessment
MAP	Risk mapping for AI products
MEASURE	Copilot effectiveness studies
MANAGE	Safety filters, content moderation

NIST 800-53 (Federal)

For federal contractors:

• Review FedRAMP status (GitHub Enterprise Cloud)
• Document AI tool authorization
• May require additional controls for classified work

---

Module 4.4: Policy Framework Design (20 min)

4.4.1 — Recommended Policy Template (10 min)

GitHub Copilot Security Policy Template

# GitHub Copilot Acceptable Use & Security Policy

**Version:** 1.0
**Effective Date:** [Date]
**Owner:** Information Security
**Review Cycle:** Quarterly / Per Major Release

## 1. Purpose
This policy defines security requirements for GitHub Copilot usage
within [Organization Name].

## 2. Scope
All employees with GitHub Copilot licenses.

## 3. Policy Requirements

### 3.1 Approved Use Cases
- Development of [approved project types]
- Code completion and assistance
- Test generation
- Documentation generation

### 3.2 Prohibited Use Cases
- Classified or restricted code
- Code under NDA without vendor approval
- Generation of security-critical algorithms without review
- Processing of PII/PHI without authorization

### 3.3 Required Controls
| Control | Setting | Enforcement |
|---------|---------|-------------|
| Public code filter | Enabled | Enterprise policy |
| Content exclusion | [Patterns] | Enterprise policy |
| Agent Mode | [Enabled/Disabled] | Enterprise policy |
| Coding Agent | [Enabled/Disabled] | Enterprise policy |

### 3.4 Review Requirements
- All Copilot-generated code subject to standard code review
- Security-sensitive code requires security review
- No automatic commits of Copilot suggestions

## 4. Compliance
Violations subject to [standard disciplinary process].

## 5. Exceptions
Exceptions require CISO approval.

## 6. Review History
| Version | Date | Changes |
|---------|------|---------|
| 1.0 | [Date] | Initial policy |

---

4.4.2 — Ongoing Security Review Process (10 min)

Release Review Cadence

flowchart LR
    A["GitHub Release<br/>(~Monthly)"] --> B["Security Review<br/>(2-5 days)"]
    B --> C{"New Features?"}
    C -->|Yes| D["Risk Assessment"]
    C -->|No| E["Confirm Controls"]
    D --> F["Update Policy"]
    E --> F
    F --> G["Communicate to Admins"]

Loading

New Feature Review Checklist

For each new Copilot feature:

• What data does it access?
• Does it respect content exclusion?
• Does it require new permissions?
• What audit events are generated?
• Is it enabled by default?
• Should we enable, disable, or defer?

Security Incident Response

Scenario	Response
Suspected data leak via Copilot	Disable for affected users, audit logs, investigate
Vulnerability in Copilot extension	Update extension, communicate to users
GitHub security incident	Monitor GitHub status, review audit logs
Malicious code suggestion accepted	Standard incident response, add to test cases

---

Track Assessment

Security Review Deliverables

After completing this track, you should produce:

• Risk assessment document (Module 1.3 template)
• Control configuration recommendations (Module 2.2)
• Compliance mapping relevant to your frameworks (Module 3)
• Acceptable use policy (Module 4.1 template)
• Ongoing review process definition (Module 4.2)

Handoff to Administrators

Provide Admin team with:

1. Approved policy settings
2. Content exclusion patterns
3. Feature enable/disable decisions
4. Audit requirements
5. Escalation procedures

---

Appendix: Security Resources

GitHub Resources

• GitHub Copilot Trust Center
• GitHub Security Documentation
• GitHub Compliance Reports

Industry Frameworks

• NIST AI RMF
• NIST CSF 2.0
• SOC 2 Trust Services Criteria

---

Version History

Version	Date	Changes
1.0	December 2025	Initial track

---

Related Documents

Document	Relevance
05 - Administrators Track	Policy configuration how-to
06 - Integration Guide	SIEM integration details
09 - Governance Templates	Ready-to-use policy templates
10 - Quick Reference Cards	Security checklist card
12 - Implementation Risks	Shadow Copilot detection
14 - Architecture & Security	Advanced security patterns