Search the Atlas

Search risks, controls, and glossary terms

HighOversight GapDemonstrated

Missing human oversight

Governance, Accountability & Compliance

Description

Material decisions are made or communicated by the model without effective human review for high-impact flows.

Example scenario

AI-drafted adverse-action letters go out with no human review.

Real-world evidenceDemonstrated

AI-specific regulatory reporting obligations (under EU AI Act, EBA stress-testing, FRB model inventories) are newly mandated and institutions are in early compliance phases; confirmed enforcement actions for failure to meet AI reporting obligations specifically—as opposed to broader model risk reporting—are still accumulating rather than an established pattern.

No public incident on record — evidence level: Demonstrated

Primary mitigations

  • Human-in-the-loop for high-impact decisions
  • approval gates
  • reviewer training
  • override authority.

Detection signals

Human-review coverage on high-impact flows; override-rate monitoring.

Mitigating controls

4
Non-agentic controls

Related risks in Governance, Accountability & Compliance