Book a demo
Topics · GOVERNANCE-AGENTS

Insurance Agent Identity NAIC | Moring

NAIC expectations translate directly into agent identity requirements for insurers. The Model Bulletin asks for a written AI Systems Program, named accountability, risk-based evaluation, and documentation retained for examination. Scoped identities for claims, underwriting, and service agents, each with an owner and per-call records, are how carriers demonstrate those expectations rather than merely describing them.

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

Short answer

NAIC expectations translate directly into agent identity requirements for insurers. The Model Bulletin asks for a written AI Systems Program, named accountability, risk-based evaluation, and documentation retained for examination. Scoped identities for claims, underwriting, and service agents, each with an owner and per-call records, are how carriers demonstrate those expectations rather than merely describing them.

How it works

The NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers sets four expectations that map cleanly onto agent identity: a written AI governance program, risk-based evaluation of AI systems, defined accountability through named owners, and documentation retained for regulatory review. The bulletin does not mention agents or tool calls explicitly, so applying it to agents is an interpretation the carrier must document, but agents acting in claims, underwriting, and customer communications fall within the scope the model describes wherever a state has adopted it.

Accountability is the expectation identity controls answer most directly. The bulletin expects named owners for AI systems and their outcomes, and identity governance research echoes it: every agent should carry a human owner, a business purpose, and a lifecycle state so review and offboarding follow the same governance path as other identities. In a carrier, that means the claims summarization agent, the quoting agent, and the fraud triage agent each appear in the inventory with an accountable person, not as features of a platform with no owner of record.

Risk-based evaluation also has an identity dimension. The bulletin asks for evaluation proportionate to consumer impact, and agent governance practice classifies actions by the damage they could do. Identity controls make that classification enforceable: scoped credentials per agent, tool grants limited to the agent's declared task, and escalation to human approval when an action affects a claim outcome, a coverage position, or a customer communication. An agent whose identity cannot be scoped cannot have proportionate controls applied, so identity scoping is a precondition for the risk-based approach the bulletin describes.

Documentation is where identity evidence earns its place. Industry analysis of NAIC examination preparation emphasizes reconstructability: can the insurer show one decision, a claim or a coverage determination, end to end, with approvals and records. Per-call identity records, which agent invoked which tool or model, under what policy, with what outcome, are the raw material for that reconstruction. Because those records are generated as enforcement happens, the carrier can answer examination questions about how it develops, deploys, and uses AI systems without rebuilding history from logs that were never designed for agents.

The adoption landscape shapes urgency. About 25 United States jurisdictions had adopted the bulletin as of April 2026, with several large states maintaining their own insurance-specific AI guidance, so the operative document for any carrier is the version its domiciliary and licensing states have adopted. Identity controls travel well across those versions because they are architectural rather than jurisdiction-specific: inventory, ownership, scoping, and per-call records satisfy the documentation expectation wherever it is tested. The EU AI Act and DORA reinforce the same traceability and resilience themes for carriers writing business in European markets.

Carriers preparing for examination should sequence the work deliberately. Inventory every agent and the tools and models it can reach. Assign an owner and a lifecycle state to each. Scope credentials and tool grants to the declared task, with expiry. Record enforcement decisions as they occur and map them to program elements in the written AIS Program. That sequence converts the bulletin's structural expectations into operational evidence tied to real identities.

Frequently asked questions

How do agent identity controls support a carrier's written AIS Program under NAIC expectations?

The bulletin expects a written program with defined accountability, risk-based evaluation, and retained documentation. Identity controls operationalize each element: every agent holds a distinct non-human identity with a named owner and lifecycle state, credentials are scoped to the declared task, and per-call records capture the policy applied and the outcome. That turns the program from description into evidence an examiner can trace to specific actors.

Does the NAIC bulletin cover AI agents even though it does not name them?

The bulletin addresses decisions or actions made or supported by AI systems, so an insurer can reasonably apply its program, risk control, third-party, and documentation expectations to agents. Because the bulletin does not mention tool calls, that mapping is an interpretation the carrier should document. Agents acting in claims, underwriting, and customer communications fall within the described scope wherever a state has adopted the bulletin.

What identity evidence best answers NAIC examination questions about agent use?

Reconstructability is the central test: show one decision, such as a claim or coverage determination, end to end with approvals and records. Per-call identity records, which agent invoked which tool, under what policy, with what result, provide that reconstruction directly. Confirm that enforcement decisions are recorded as they occur and that agent permissions are scoped, expired, and revoked with named approvers for changes.

How does the state-by-state adoption of the bulletin affect identity planning?

Adoption varies, so the operative document for a carrier is the version its domiciliary and licensing states have adopted, with roughly 25 jurisdictions on board as of early 2026. Identity controls travel well across versions because they are architectural: inventory, ownership, scoping, and per-call records satisfy the documentation expectation in each adopting state without jurisdiction-specific redesign of the agent identity layer.

Sources

  1. content.naic.org
  2. www.vijil.ai
  3. trussed.ai
  4. www.moring.ai
  5. compassmsp.com
  6. - artificialintelligenceact.eu
  7. - eur-lex.europa.eu