Book a demo
Topics · GOVERNANCE-AGENTS

Insurance Agent Registry NAIC | Moring

The NAIC frames AI oversight for insurers through model guidance and its artificial intelligence topic work: a written AI systems program, governance and risk controls proportionate to the use case, and documentation an examiner can review. An agent registry operationalizes those expectations for autonomous agents, giving each claims, underwriting or servicing agent a recorded owner, permission boundary and evidence trail that examiners can inspect.

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

Short answer

The NAIC frames AI oversight for insurers through model guidance and its artificial intelligence topic work: a written AI systems program, governance and risk controls proportionate to the use case, and documentation an examiner can review. An agent registry operationalizes those expectations for autonomous agents, giving each claims, underwriting or servicing agent a recorded owner, permission boundary and evidence trail that examiners can inspect.

How it works

The National Association of Insurance Commissioners treats artificial intelligence as a standing supervisory topic, with public material describing how AI is used across underwriting, pricing, customer service, claims handling, marketing and fraud detection, and how insurers analyze policyholder data to identify patterns and support decision-making. That framing matters for agent governance: the workflows where agents act are the same ones the association's material identifies as AI-bearing, so a carrier's registry should map directly onto them, one entry per claims triage, underwriting support or servicing agent, rather than a parallel inventory that examiners must reconcile.

The anchor instrument is the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, adopted December 4, 2023. It is guidance rather than a model law: it restates that existing prohibitions on unfair trade practices, unfair discrimination and unfair claims settlement apply in full when a decision is made or supported by an AI system, and it describes an AI Systems Program with senior management and board accountability, risk controls, testing for errors, bias and unfair discrimination, and third-party oversight where the insurer remains responsible. Adoption proceeds state by state, so carriers confirm the posture in each jurisdiction where they write business.

Adoption breadth explains the urgency. Sources tracking the rollout report that twenty-five states plus the District of Columbia had adopted the bulletin, with other counts citing twenty-six jurisdictions including states with their own insurance AI rules, and at least twenty-eight jurisdictions having enacted the related Insurance Data Security Model Law. For a carrier operating across state lines, the practical reading is that AI governance expectations are functionally national: the registry, the evidence it holds and the program it supports should be built to the demanding end of the adoption spectrum rather than to the most permissive state.

The bulletin's implementation challenge is the operational layer. Governance program guides describing state reviews consistently find that a policy document alone is rarely sufficient: examiners look for a person or committee responsible, a record of decisions made under the program, and controls that trace back to specific AI use cases in production. Cross-functional committees drawing on legal, compliance, actuarial, technology and the owning business unit are the mechanism; single-department ownership is a recurring finding against carriers. The registry is what makes the committee's oversight traceable, because every use case decision resolves to a registered entry.

Third-party accountability is the element carriers most often underestimate, and registration is where it becomes concrete. The bulletin makes the insurer accountable for AI sourced from a third party, which means vendor-supplied agents embedded in claims platforms enter the same registry as internal builds, with vendor lineage recorded. The association's public topic material also emphasizes the consumer-facing stakes, from chatbots answering policyholder questions to claims processing that estimates repair costs; an agent in any of those paths that cannot produce its owner, its approved scope and its decision record leaves the carrier exposed in exactly the review the bulletin anticipates.

For agent governance specifically, the NAIC framework converts into registry fields. The written AI program becomes the policy layer; the governance committee becomes the approval authority recorded on each entry; proportionate risk controls become tiering by customer impact; documentation becomes the sealed decision records, model lineage and permission boundaries attached to each agent. A carrier that operates this way can answer the supervisory question the control literature summarizes, dated evidence for a specific system of what it was allowed to do, what it did and who checked it, from the registry itself rather than from reconstruction under examination pressure.

Frequently asked questions

What does the NAIC Model Bulletin expect from insurers using agents?

It expects a written AI systems program covering every lifecycle phase, including third-party systems, with senior management and board accountability, risk controls proportionate to the use case, testing for errors, bias and unfair discrimination, and documentation an examiner can review. It also confirms that existing prohibitions on unfair trade practices and unfair discrimination apply in full to AI-supported decisions. For agents, this means each claims, underwriting or servicing agent needs a recorded owner, approval and evidence trail.

Is the NAIC bulletin binding law on insurers?

It is guidance, not a self-executing model law; authority rests with each state insurance regulator that adopts it. Adoption has been broad, with counts in the industry literature of twenty-five states plus the District of Columbia and twenty-six jurisdictions including states with their own AI rules. The practical posture for a multi-state carrier is to build to the demanding end of the adoption spectrum and confirm the current posture in each jurisdiction where it writes business, rather than treating the bulletin as optional.

How does the registry support a NAIC-aligned program?

The registry turns program expectations into traceable records: each agent entry carries its owner, the committee approval for its current build, its permission boundary, model lineage and its decision evidence. That gives examiners what program reviews consistently look for, a record of decisions made under the program and controls tracing back to specific production use cases. A claims triage agent's entry, for example, answers who approved it, what it may reach and what it did, directly from the registry.

Do vendor-supplied agents fall under the bulletin's expectations?

Yes, and this is the element carriers most often underestimate: the bulletin makes the insurer accountable for AI sourced from third parties. Vendor agents embedded in claims or servicing platforms therefore belong in the same registry as internal builds, with vendor lineage recorded and the same owner, permission and evidence discipline applied. When an examiner questions a vendor agent's decision, the carrier, not the vendor, is the party that must produce the record, so registration is the carrier's protection.

Sources

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