Book a demo
Topics · REGULATORY

Insurance Model Risk Access Control | Moring

Access control is the first control that reduces model risk in insurance: a continuously refreshed AI inventory covering internal models, vendor-embedded AI, agents, and datasets, built from IAM identities, procurement records, and workflow owners rather than self-reporting. The inventory scopes every other control.

On this page
  1. Short answer
  2. How it works
  3. FAQs
  4. Related reading
  5. Sources

Short answer

Access control is the first control that reduces model risk in insurance: a continuously refreshed AI inventory covering internal models, vendor-embedded AI, agents, and datasets, built from IAM identities, procurement records, and workflow owners rather than self-reporting. The inventory scopes every other control.

How it works

Access control is the control that makes model risk manageable in insurance, and the first control is the inventory. An inventory built from self-reporting undercounts, always, so the inventory must be built from evidence: IAM identities that belong to non-human actors, procurement records for AI-flagged vendors, and workflow owners naming the systems that touch their processes. The inventory is a continuously refreshed AI Bill of Materials covering internal models, vendor-embedded AI, agents, and the datasets behind them. The effort grows with the number of business units, vendors, and adopting states involved, and the inventory is the control that scopes every other control.

The inventory is also where the classification is decided. The classification is by use case, not model: the same model is low-risk summarizing meeting notes and high-risk screening job applicants, and classifying by use case keeps oversight proportionate. For an insurance carrier, the classification is by the workflow the model touches: underwriting, pricing, claims administration, fraud detection, and policy servicing. The highest regulatory scrutiny in the stack is underwriting and risk selection, and the risk classification is the control that makes the oversight proportionate to the risk.

The inventory is also where the access control is decided. Every agent deployed within the organization must have a unique, machine-readable identity, a named human owner, a clearly documented purpose, and a record in a central registry. Shared API keys or generic service accounts for agents are a direct path to untraceable actions and audit failure. Without that identity layer, the inventory names a system nobody can hold accountable, and the access control that follows is unenforceable. Every later decision, from approval thresholds to rollback, inherits its strength from the precision of these identities.

The inventory is also where the audit trail is produced. The audit trail of agent actions and the access control decision supports post-incident review and regulatory documentation. The EU AI Act places human oversight at the center of high-risk systems, including the ability to intervene in or stop operation when needed, and the NAIC Model Bulletin, adopted by roughly 24 states, expects the same rigor applied to cybersecurity, financial controls, privacy, and operational risk to be applied to AI, with a documented AIS Program expected to show the insurer can govern what runs in production.

The failure modes are predictable. The inventory is built from self-reporting, so it undercounts, always. The inventory is built but the classification is by model, not use case, so the oversight is not proportionate to the risk. The inventory is built but the access control is not enforced, so the inventory is documentation. Or the inventory is enforced but the audit trail is not tamper-evident, so the AIS Program is a document rather than a claim. The remedy is an inventory built from evidence, a classification by use case, an access control that is enforced, and an audit trail that is tamper-evident.

Frequently asked questions

Why is the inventory the first access control for insurance model risk?

An inventory built from self-reporting undercounts, always, so the inventory must be built from evidence: IAM identities that belong to non-human actors, procurement records for AI-flagged vendors, and workflow owners naming the systems that touch their processes. The inventory is a continuously refreshed AI Bill of Materials covering internal models, vendor-embedded AI, agents, and the datasets behind them. The inventory is the control that scopes every other control, and the first pass is a matter of days rather than weeks when the queries are right.

How should an insurance carrier classify model risk for access control?

The classification is by use case, not model: the same model is low-risk summarizing meeting notes and high-risk screening job applicants, and classifying by use case keeps oversight proportionate. For an insurance carrier, the classification is by the workflow the model touches: underwriting, pricing, claims administration, fraud detection, and policy servicing. The highest regulatory scrutiny in the stack is underwriting and risk selection, and the risk classification is the control that makes the oversight proportionate to the risk.

What identity model should an insurance carrier use for agents?

Every agent deployed within the organization must have a unique, machine-readable identity, a named human owner, a clearly documented purpose, and a record in a central registry. Shared API keys or generic service accounts for agents are a direct path to untraceable actions and audit failure. The inventory is the control that makes the access control meaningful, and the inventory is the control that scopes every other control. The inventory is the control that makes the classification enforceable.

What does the NAIC Model Bulletin require of an insurance carrier's access control?

The NAIC Model Bulletin, adopted by roughly 24 states, expects the same rigor applied to cybersecurity, financial controls, privacy, and operational risk to be applied to AI, with a documented AIS Program expected to show the insurer can govern what runs in production. The inventory is the control that scopes every other control, and the inventory is the control that makes the access control meaningful. The EU AI Act places human oversight at the center of high-risk systems, including the ability to intervene in or stop operation when needed.

Sources

  1. ai.wharton.upenn.edu
  2. adaptivesecurity.com
  3. captive.com
  4. charlee.ai
  5. cranium.ai
  6. databricks.com
  7. - content.naic.org
  8. - artificialintelligenceact.eu
  9. - eur-lex.europa.eu