Skip to content

Verification & Testing: Control 2.7 - Vendor and Third-Party Risk Management

This playbook provides verification and testing guidance for Control 2.7.

Verification Checklist

Step Action Expected Result
1 Review connector inventory document Complete list of all third-party connectors
2 Check vendor security assessments Current assessments for all Tier 1/2 vendors
3 Verify contracts have security clauses Required clauses present in all vendor contracts
4 Confirm vendor access is monitored Audit logs showing connector activity
5 Test DLP policy enforcement Blocked connectors cannot be used
6 Review board reporting Quarterly vendor risk reports delivered
7 Verify incident response procedures Documented and tested for vendor issues

Vendor Assessment Requirements by Zone

Assessment Area Zone 1 Zone 2 Zone 3
Vendor Vetting Self-certification Basic questionnaire Comprehensive assessment
Security Documentation Optional SOC 2 recommended SOC 2 Type II required
Contract Review Standard terms Legal review Security addendum required
Monitoring Frequency Annual Quarterly Continuous
Audit Rights Not required Recommended Required
Exit Planning Optional Documented Tested annually
Board Reporting None Summary Detailed risk report

AI Vendor Assessment Questionnaire Sections

Section A: Model Information

  • What AI/ML models power the service?
  • How are model updates handled?
  • What is the model change notification process?
  • Is model documentation (model card) available?

Section B: Training Data

  • What data was used to train the model?
  • How is training data governance managed?
  • Does the vendor train on customer inputs?
  • Are bias testing results available?

Section C: Output Quality and Safety

  • What output quality controls are in place?
  • What safety guardrails exist?
  • How is output accuracy measured?
  • What is the incident response process for AI failures?

Section D: Transparency and Explainability

  • Can outputs be explained?
  • What audit capabilities exist?
  • Is explainability documentation available?

Section E: Compliance and Regulatory

  • AI-specific certifications held?
  • Regulatory examination support?
  • AI governance framework?

Section F: Model Risk Management Integration

  • OCC 2011-12/SR 11-7 support?
  • Model inventory information provided?

AI Vendor Risk Scoring

Factor Weight Scoring Criteria
Model transparency 20% Full disclosure (low risk) to black box (high risk)
Training data governance 15% Documented and audited (low) to undisclosed (high)
Output quality controls 20% Comprehensive controls (low) to minimal (high)
Customer data protection 15% No training on customer data (low) to unrestricted (high)
Regulatory readiness 15% AI-specific certifications (low) to no attestations (high)
Incident response 15% Defined AI incident process (low) to undefined (high)

Dynamic Tool Governance Checklist

  • Default-deny policy for runtime tool discovery
  • Plugin allowlist maintained and enforced
  • Community plugins require full security review
  • Automatic updates disabled for third-party tools
  • Transitive data exposure mapped for tool chains
  • Agent-to-agent composition risks assessed
  • Marketplace installations blocked by default
  • Tool chain audit logging enabled
  • Quarterly review of approved tool inventory

Transitive Data Exposure Mapping Template

TRANSITIVE DATA EXPOSURE MAPPING
================================
Agent: [Agent Name]

Primary Tool: [Connector Name] ([Publisher])
  └── Data Retrieved: [Data type]

Secondary Tool: [Service Name] ([Third-Party])
  └── Data Sent: [Data type]
  └── Third Party: [Vendor name]
  └── Data Residency: [Location]

Tertiary Tool: [Service Name] ([Third-Party])
  └── Data Sent: [Data type]
  └── Third Party: [Vendor name]
  └── Data Residency: [Location]

RISK ASSESSMENT:
├── Customer data flows to [X] third parties
├── Data residency: [Known/Unknown]
├── Transitive exposure disclosed: [Yes/No]
└── RECOMMENDATION: [Action]

Contract Requirements Checklist

Standard Clauses

  • Data protection and encryption requirements
  • Incident notification (24 hours for critical)
  • Audit rights (annual minimum)
  • Subprocessor approval requirements
  • Data return/destruction on termination

AI-Specific Clauses

  • Model change notification
  • Training data requirements
  • Output monitoring rights
  • Human oversight provisions
  • Explainability requirements

Updated: January 2026 | Version: v1.2