CSOAI   Home · Journal · Certification · Fabric
The 52-Article Charter · 22 of 52 · full text

Article 22: Cybersecurity Requirements

Published from the canonical CSOAI Partnership Charter (effective 15 January 2026). Full text below.

Version: 1.0 Effective Date: January 15, 2026, 09:00 GMT Status: Technical Article - Security Standards

PREAMBLE

This Article establishes comprehensive cybersecurity requirements for all AI systems. AI systems are high-value targets for adversaries—state actors, cybercriminals, competitors. A compromised AI system can cause catastrophic harm. Security is not optional. Security is survival.

Core Principle: Defense in depth. Assume breach, minimize impact.

22.1 SECURITY FRAMEWORK REQUIREMENTS

22.1.1 NIST Cybersecurity Framework (CSF 2.0)

Mandatory for All AI Systems:

The NIST CSF provides a flexible framework for managing cybersecurity risk. All CSOAI-licensed AI must implement the six core functions:

GOVERN:

IDENTIFY:

PROTECT:

DETECT:

RESPOND:

RECOVER:

CSOAI Requirements by Risk Tier:

| Risk Tier | CSF Implementation Level |
|-----------|-------------------------|
| Low | Partial (Tier 1) - Inform decisions with risk awareness |
| Medium | Risk-Informed (Tier 2) - Informed risk management |
| High | Repeatable (Tier 3) - Consistent risk management approach |
| Critical | Adaptive (Tier 4) - Continuously improving, data-driven |
Documentation:

22.1.2 ISO/IEC 27001:2022 Certification

Information Security Management System (ISMS):

Required for Medium+ Risk Tiers:

ISO 27001 Requirements:

Annex A Controls (114 total):

All applicable controls implemented, including:

Statement of Applicability (SoA):

22.1.3 SOC 2 Type II Audit

Service Organization Control:

Required Annually for High/Critical Tiers:

Trust Service Criteria:

Security:

Plus (if applicable):

Report Distribution:

22.1.4 Zero Trust Architecture

"Never Trust, Always Verify":

Principles:

Implementation:

Identity-Centric Security:

Device Security:

Network Segmentation:

Data-Centric Security:

NIST SP 800-207 Compliance:

22.2 THREAT PROTECTION

22.2.1 Endpoint Detection and Response (EDR)

All Endpoints Must Have EDR:

Capabilities Required:

Approved Solutions:

Configuration:

Coverage:

22.2.2 Extended Detection and Response (XDR)

For High/Critical Systems:

Beyond EDR - Correlation Across:

Benefits:

Implementation:

22.2.3 Security Information and Event Management (SIEM)

Centralized Log Management and Analysis:

Required for Medium+ Tiers:

Capabilities:

Approved Solutions:

Log Sources:

Use Cases:

24/7 Monitoring:

22.2.4 Intrusion Detection and Prevention Systems (IDS/IPS)

Network-Based Threat Detection:

IDS (Detection):

IPS (Prevention):

Deployment:

Approved Solutions:

Tuning:

22.2.5 Web Application Firewall (WAF)

For AI Systems with Web Interfaces:

Protection Against:

Deployment:

OWASP Top 10 Coverage:

Configuration:

Special Considerations for AI:

22.2.6 DDoS Protection

Distributed Denial of Service Mitigation:

Layers of Protection:

Layer 3/4 (Network/Transport):

Layer 7 (Application):

Always-On Protection:

Capacity:

22.3 VULNERABILITY MANAGEMENT

22.3.1 Vulnerability Scanning

Automated and Continuous:

Tools:

Scanning Frequency:

| Asset Type | Scan Frequency |
|-----------|---------------|
| Production systems | Weekly |
| Development/staging | Monthly |
| Cloud infrastructure | Daily (automated) |
| Containers/images | Every build |
| Dependencies | Daily (automated) |
Scan Scope:

Vulnerability Databases:

22.3.2 Patch Management

Timely Remediation Required:

Patching SLAs (from vulnerability disclosure):

| Severity | Patching Deadline | Justification if Delayed |
|----------|------------------|------------------------|
| Critical (CVSS 9.0-10.0) | 7 days | Requires C-level approval |
| High (CVSS 7.0-8.9) | 30 days | Risk assessment required |
| Medium (CVSS 4.0-6.9) | 90 days | Standard priority |
| Low (CVSS 0.1-3.9) | 180 days | As resources permit |
Process:

Change Management:

Virtual Patching:

22.3.3 Penetration Testing

Simulated Attacks to Identify Weaknesses:

Frequency:

| Risk Tier | Penetration Test Frequency | Type |
|-----------|--------------------------|------|
| Low | Every 2 years | Automated scan acceptable |
| Medium | Annual | Manual testing (limited scope) |
| High | Annual | Comprehensive manual testing |
| Critical | Biannual (every 6 months) | Red team exercise |
Scope:

Methodology:

Rules of Engagement:

Reporting:

External Testers:

22.3.4 Bug Bounty Programs

Crowdsourced Security:

Encouraged for All, Required for Critical:

Platforms:

Scope:

Rewards:

| Severity | Bounty Range |
|----------|-------------|
| Critical | $5,000 - $50,000 |
| High | $1,000 - $10,000 |
| Medium | $500 - $2,000 |
| Low | $100 - $500 |
| Informational | Swag/recognition |
Response SLA:

Benefits:

22.4 INCIDENT RESPONSE

22.4.1 Security Incident Response Plan (SIRP)

Documented and Tested Plan:

Required for All Risk Tiers:

Plan Contents:

1. Preparation:

2. Identification:

3. Containment:

4. Eradication:

5. Recovery:

6. Lessons Learned:

Testing:

22.4.2 Security Operations Center (SOC)

24/7 Monitoring and Response:

Required for Critical Risk Tier:

SOC Functions:

For High Risk:

For Medium Risk:

SOC Tools:

22.4.3 Incident Classification

Severity Levels:

Critical:

High:

Medium:

Low:

Response Times:

| Severity | Initial Response | Full Containment | Resolution |
|----------|-----------------|-----------------|------------|
| Critical | 15 minutes | 4 hours | 24 hours |
| High | 1 hour | 24 hours | 1 week |
| Medium | 4 hours | 1 week | 1 month |
| Low | 1 business day | As needed | As needed |

22.4.4 Notification Requirements

When Security Incidents Occur:

Internal Notification:

External Notification:

CSOAI (Article 13.4):

Regulators (if applicable):

Customers/Users:

Law Enforcement:

Public Disclosure:

22.4.5 Post-Incident Review

Always Conduct After Every Incident:

Timeline: Within 2 weeks of incident resolution

Participants:

Review Questions:

Outputs:

22.5 SUPPLY CHAIN SECURITY

22.5.1 Software Bill of Materials (SBOM)

Inventory of All Components:

Required for Medium+ Risk Tiers:

SBOM Contents:

Formats:

Generation:

Usage:

Sharing:

22.5.2 Dependency Management

Third-Party Libraries and Packages:

Risks:

Best Practices:

Dependency Pinning:

Automated Scanning:

Private Registry:

Minimal Dependencies:

License Compliance:

22.5.3 Vendor Security Assessments

Third-Party Software and Services:

Due Diligence Before Procurement:

Security Questionnaire:

Risk-Based Assessment:

Ongoing Monitoring:

Contract Requirements:

22.5.4 Open Source Security

Using Open Source Components:

Benefits:

Risks:

Risk Mitigation:

Vetting:

Monitoring:

Contribution:

Alternatives:

22.5.5 Hardware Supply Chain

For AI Training Infrastructure:

Risks:

Mitigations:

Critical Systems:

22.6 CONCLUSION

Cybersecurity is not a destination, it's a journey. Threats evolve, defenses must too. CSOAI requires continuous improvement, not one-time compliance.

The stakes are high:

Defense in depth is essential:

No single control is sufficient. Layers provide resilience.

Culture of security:

CSOAI supports members:

Together, we secure AI. Individually, we struggle.

Effective Date: January 15, 2026, 09:00 GMT "Secure the Foundation, Protect the Future"

REFERENCES

NIST. (2024). NIST Cybersecurity Framework (CSF) 2.0. NIST.

ISO/IEC. (2022). ISO/IEC 27001:2022 - Information Security Management Systems.

AICPA. (2024). SOC 2 Trust Services Criteria.

NIST. (2020). NIST SP 800-207 - Zero Trust Architecture.

OWASP. (2023). OWASP Top 10 - 2023.

MITRE. (2024). ATT&CK Framework for Enterprise.

CISA. (2024). Known Exploited Vulnerabilities Catalog.

NTIA. (2021). Software Bill of Materials (SBOM) - Framing Working Group.

END OF ARTICLE 22

Next: Article 23 - Model Development Standards (FULL VERSION)

From charter to certificate. This article is part of the standard behind Watchdog Certification — independent assessment, Ed25519-signed, publicly verifiable. The crosswalks to the EU AI Act, ISO/IEC 42001 and 18 more frameworks are in the Crosswalk Library; the runtime tools are in the fabric.

The 52-Article Charter is published in full in the Journal. Bespoke briefings: hello@meok.ai.