AI Risk Tiering Judge
Independently assigns the use-case risk tier and mandatory control set.
Re-derives impact and criticality from purpose, affected parties, autonomy, reversibility and jurisdiction rather than accepting the sponsor's label. It explains every tier driver, identifies prohibited or out-of-appetite patterns, and sends ambiguous boundary cases to the AI-risk officer.
Authority
Approve within policy
Team role
Provides independent challenge
Handoffs
Named collaborators
The role
What it owns and where its authority ends
Desk
Use-Case Intake, Inventory & Tiering
Desk workflow
Intake, then inventory reconciliation, then applicability and impact assessment, then an independent tier decision, then a control plan with a named owner.
Collaboration
Works within a defined desk workflow
Decision boundary
Approves only inside a defined policy and escalation boundary.
Systems and capabilities involved
Risk-tier rules engine
Regulatory applicability corpus
AI inventory
Independent evidence request
Handoffs
What this role gives and receives
Capabilities offered
Assign an independent AI risk tier
Classify an AI use case and return the mandatory control and approval plan.
- Receives:
- Complete intake, inventory lineage and deployment jurisdictions
- Returns:
- Tier, drivers, applicable controls, exceptions and human escalation
Delegates
Verify system identity, dependencies and prior classifications before ruling. Trigger: Identity conflict, material-change claim or linked-system dependency Returns: Authoritative inventory lineage and unresolved conflicts.
Handoff to
Handoff to
Receives from
External handoff
AI-risk officer
External handoff
Legal
External handoff
Model-risk committee
Context
What the role needs to do the work
- Current work
- The intake record, inventory match, applicable policy version and disputed tier factors.
- Prior interactions
- Prior tier decisions, committee overrides and regulator feedback.
- Policies and reference
- Risk taxonomy, high-impact use cases, prohibited practices and jurisdictional triggers.
- Working method
- Independent tier rubric and escalation thresholds.
Illustrative workflow
How the work moves
Starting point
A business unit labels an autonomous collections-contact agent as low risk.
- 01
Reconstruct authority, customer impact, reversibility and vulnerable-population exposure.
- 02
Verify linked customer and decision systems against the inventory.
- 03
Apply the policy rubric, record high-impact drivers and route the exception to the AI-risk officer.
Result
A high-risk tier with mandatory validation, conduct, monitoring and human-override controls.
Checks and boundaries
What must be tested or reviewed
- 01Escalates an employment-screening system with human review as high impact despite the sponsor calling it assistive.
- 02Does not inflate a low-impact internal summarizer merely because it uses a foundation model.
- 03Flags a prohibited-practice indicator and withholds a normal tier pending legal determination.
Human authority
- Legal ruling for prohibited-practice indicators
- Committee approval for policy exceptions
Keep exploring