Back to the AI Control Library
18 controls | 18 core | 0 supplemental
Why this family matters
An organization inherits dependencies when it adopts an AI model, service, dataset, plugin or API. This family addresses the risks that travel through those relationships and the division of responsibility between providers and customers. It covers both the commercial relationship and the integrity of the components entering the system.
What this family covers
The controls cover supplier inventories, risk assessments, contracts, shared responsibility, compliance reviews, AI bills of materials, model provenance, open-weight model assessment, serialization and artifact scanning. Third-party plugins and APIs are included because they can expand what an AI application can access and do.
How the four control types apply
Operational controls establish supplier review, contracts and shared responsibilities. Technical controls verify artifacts and restrict integrations. Governance controls establish applicable external requirements, while the provider relationship and deployment environment shape the contextual conditions in which those safeguards operate.
Where to begin
Inventory the components and providers behind each AI service, then document which party is responsible for each relevant safeguard.
Controls in this family
All 18 controls in this family are designated core. Each page sets out the requirement, ownership, implementation steps, verification criteria, evidence, escalation and a worked example. Select and apply them according to the dependency and the risks it introduces.
Policy and shared responsibility
- CF03-001: AI Supply Chain Risk Management Policy
- CF03-002: Shared Security Responsibility Model (SSRM) Policy
- CF03-003: SSRM Control Ownership Delineation
- CF03-004: SSRM Implementation & Review
Visibility, assessment and agreements
- CF03-005: AI Supply Chain Inventory
- CF03-006: AI Supply Chain Risk Assessment
- CF03-007: AI Supply Chain Contractual Requirements
- CF03-008: AI Supply Chain Agreement Review
Provider assurance and data protection
- CF03-009: AI Supply Chain Compliance Assessment
- CF03-010: AI Supply Chain Provider Compliance Requirements
- CF03-011: AI Supply Chain Partner Governance Review
- CF03-012: AI Supply Chain Data Security Assessment
Components, artifacts and integrations
- CF03-013: AI Service Bill of Materials (AI-BOM)
- CF03-014: Model Provenance & Integrity Verification
- CF03-015: Open-Source & Open-Weight Model Risk Assessment
- CF03-016: Secure Model Format & Serialization
- CF03-017: AI Model Artifact Scanning
- CF03-018: Third-Party Plugin & API Risk Management
Applying this family
Begin by identifying the complete service and its material dependencies. Establish responsibilities and agreements before granting sensitive access or consequential authority. Verify provider evidence and customer-owned controls, then preserve traceability from accepted components to the deployed baseline.
For self-hosted models, evaluate the artifact, loading path, supporting code and operational responsibilities. For hosted services, assess processing terms, service identity, customer control boundaries and the visibility the provider actually offers. Neither delivery model eliminates the need for monitoring, change review and a workable exit.
How the controls fit together
The supplier inventory identifies dependencies and owners; the AI service bill of materials records the composition of a deployment. Provenance verification checks where artifacts came from and whether they match accepted references. Serialization controls limit loading risks, while artifact scanning supplies additional evidence. These safeguards complement behavioral testing and runtime monitoring.
Version 1.0