AI Governance Global/Local Feature Guide
Overview
Establish standardized global records for AI systems, models, and datasets while maintaining contextual variations to enable their use across business units, use cases, and lifecycle stages.
This capability ensures consistent governance practices across the enterprise while accommodating variations in regulatory frameworks, risk profiles, and needs across the AI lifecycle. Teams can track AI asset evolution through a clear audit trail, enforce standardized attributes at the global level, and manage jurisdiction-specific or use-case-specific requirements at the local level. This approach streamlines compliance, reduces redundancy, and provides clear lineage of how AI assets are used throughout the enterprise.
Solving Complex Problems
Organizations face multiple challenges in governing AI assets across complex enterprises:
-
Inconsistent governance practices when the same AI model or dataset is used across different business units and jurisdictions, leading to compliance gaps and inefficiencies
-
Difficulty tracking and managing varying risk profiles / controls when an AI asset is used for different use cases
-
Limited visibility into how AI assets evolve through different stages of the AI lifecycle from pre-deployment to post-deployment retraining, making it difficult to maintain auditable records
Example Use Case
A foundation model (e.g. GPT-4) is registered globally with core metadata, like description, provider, hosting location; assessment results; risks and controls; related records, documents, etc.. A regional team creates a linked model record to reflect their use of GPT-4, inheriting most metadata and relate but overriding business owner, purpose, and local risk.
Considering this scenario, the following user types have their own respective objectives in order preserve and maintain locality across a global business:
- As a global governance lead
-
, I want to define reusable AI inventory records that local teams can instantiate without losing consistency.
- As a local governance user,
-
I want to modify relevant fields (e.g., local approvers) without breaking the link to the global record.
- As a risk/compliance analyst
-
, I want to trace where my risks or assessments have been reused or overridden.
To configure global/local inheritance settings
To set up inventory record as the global record
Only Model or AI system records can be set up as global records. The configuration is the same for both inventory types.
To link local records to global records
Only Model or AI system records can be linked to global records. The configuration is the same for both inventory types.