Insurance Regulatory Governance
NAIC AI Model Governance Consulting for Insurers
What AI model governance informed by the NAIC bulletin actually means for a carrier building underwriting, claims, and pricing systems, and why the Model Bulletin's principles-based language leaves most of the hard implementation work to the insurer.
Kriv AI provides AI model governance consulting for insurers, informed by the NAIC bulletin: an outside team that inventories every AI system touching underwriting, claims, pricing, and marketing, builds the governance program a state insurance department's AI Model Bulletin exam expects, and validates model outputs against unfair-discrimination and fairness standards before an examiner asks for it.
context
What Is NAIC Model Governance for AI?
NAIC model governance for AI is the written program an insurer keeps to show that the AI systems and predictive models it uses are governed responsibly. It comes from the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, which NAIC members adopted on December 4, 2023. The bulletin expects an insurer to maintain a written program, which it calls an AIS Program, covering governance, risk management controls and internal audit. It is a model document: each state insurance department decides whether to issue it. Whatever a state does, the insurer's duty is to follow existing insurance law, including the rules on unfair trade practices and unfair discrimination.
provides naic ai
Who Provides NAIC AI Model Governance Consulting for Insurers
Kriv AI provides AI model governance consulting for insurers, informed by the NAIC bulletin: an outside team that inventories every AI system touching underwriting, claims, pricing, and marketing, builds the governance program a state insurance department's AI Model Bulletin exam expects, and validates model outputs against unfair-discrimination and fairness standards before an examiner asks for it.
The work centers on the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, adopted by NAIC membership in December 2023. Check your state's status. It is not a single federal rule. Each state's department of insurance decides whether and how to issue it, so a carrier licensed in multiple states is really managing a family of closely related state bulletins rather than one national standard.
The bulletin's core expectation is straightforward to state and hard to operationalize: any decision or action affecting a consumer, taken or supported by a predictive model, algorithm, or AI system, has to comply with existing unfair trade practices and unfair discrimination law, and the insurer has to be able to show a regulator how it governs that. That is the gap this practice fills, turning the bulletin's principles into an inventory, a testing protocol, a governance structure, and documentation that survives a market conduct exam.
naic model governance
Why NAIC Model Governance Is So Hard to Implement for AI
NAIC model governance is hard to implement for AI because the bulletin is deliberately principles-based rather than a technical checklist, because it is adopted state by state with small variations instead of once nationally, and because the systems it covers span underwriting, claims, and marketing teams that have never shared a common model inventory or a common definition of what counts as AI.
The bulletin tells an insurer to have a governance framework proportionate to the risk of its AI use, but it does not specify a bias-testing method, a documentation template, or a validation cadence the way a technical standard would. That is intentional, so it can cover everything from a simple rules-based rating factor to a generative underwriting assistant, but it means two carriers can both be technically compliant with very different levels of actual rigor, and an examiner has real discretion in what they consider adequate.
State-by-state adoption compounds this. A multi-state carrier is tracking which departments have issued the bulletin, whether any have layered on their own examination guidance, and keeping one governance program that satisfies all of them at once rather than rebuilding it per state.
The organizational problem is usually the bigger one. Actuarial and pricing teams, claims operations, marketing, and compliance each run their own models and vendor tools, often without a shared inventory. A model governance program has to reach across all of those groups, assign accountable ownership at the senior management level, and get third-party and vendor-built models (which insurers frequently cannot fully inspect) into the same inventory and testing regime as models built in-house. Most of the effort in a real engagement goes into that cross-functional coordination, not into the regulatory reading. For the claims workflow in particular, see our page on AI claims handling with adjuster oversight. For an actuarial example, see our design for actuarial reserving software on Databricks.
context
What Does the Model Bulletin Expect an AIS Program to Cover?
The bulletin sets out its guidance for an AIS Program under four headings: general guidelines, governance, risk management and internal controls, and third-party AI systems and data. Governance means senior management accountable to the board or a board committee, with a structure that can include actuarial, data science, underwriting, claims, compliance and legal. Risk management and internal controls cover an inventory of predictive models, documentation, data practices, validation and retesting, and protection of consumer information. Third-party AI and data call for due diligence on vendors and, where appropriate, contract terms on audit rights and cooperation with regulators. Among the general guidelines, controls should be proportionate to the potential harm to consumers. The program may sit inside the insurer's enterprise risk management program, and may draw on a framework such as the NIST AI Risk Management Framework.
insurers implement naic
How Insurers Implement NAIC AI Governance, Step by Step
Insurers implement NAIC AI governance by working through a fixed sequence: build a complete AI system inventory, assign risk tiers, put named accountable owners and a governance committee in place, define testing and validation appropriate to each tier, document the program for exam readiness, and monitor it on an ongoing basis rather than treating it as a one-time project.
1. Build the AI system inventory
Catalog every model, algorithm, and AI system that touches a consumer-facing decision: underwriting and rating, claims triage and fraud scoring, marketing and lead scoring, and any vendor or third-party tool used in those processes. Vendor-built models go in the inventory too, even when the insurer cannot see inside them.
2. Risk-tier each system
Rank systems by how directly they affect a consumer outcome and how much human review sits between the model's output and the final decision. A fully automated adverse underwriting decision sits in a higher tier than a marketing recommendation a human reviews before acting on it.
3. Assign accountable ownership
Put senior management in charge of the program, accountable to the board or a board committee, and form a cross-functional governance committee spanning actuarial, data science, underwriting, claims, compliance, IT, and legal, so no system sits in a gap between departments.
4. Define testing and validation by tier
Set the testing protocol for each risk tier: fairness and unfair-discrimination testing against protected classes and their proxies, output accuracy checks, and, where the system is a large language model rather than a traditional predictive model, structured output review and red-teaming instead of classical backtesting.
5. Document for exam readiness
Write the governance framework, inventory, and testing results into the form a market conduct exam expects: what the system does, who owns it, how it was tested, and what the results showed, in language a non-technical examiner can follow.
6. Monitor and re-certify
Re-test higher-tier systems on a set cadence, re-certify the inventory when a system changes or a new one is deployed, and keep the governance committee meeting on a schedule so the program does not go stale between exams.
responsible ai means
What Responsible AI Means for Insurance Underwriting
Responsible AI for insurance underwriting means the rating variables and model logic behind an underwriting decision have been tested for proxy discrimination, an adverse decision can be explained to the consumer and to a regulator in plain language, a human reviews or can override automated declines and non-standard pricing, and every step of that process is documented as it happens rather than reconstructed after a complaint.
In practice this starts with the rating variables themselves. A variable that is facially neutral, such as a credit-based score, zip code proxy, or device or browsing signal, can still correlate with a protected class closely enough to function as an unlawful proxy. Responsible underwriting governance tests for that correlation directly rather than assuming a variable is safe because it does not name a protected class.
It also means the model's reasoning is explainable at the level an adverse-action notice requires: an applicant who is declined or rated up has to be able to learn why in terms tied to their own application, not a model score with no underlying explanation. For any system where the underwriting decision is fully automated, a documented human review or override path has to exist for the higher-risk outcomes, and the governance program has to be able to show, after the fact, that the review actually happened.
context
What Can a Regulator Ask an Insurer About Its AI?
Section 4 of the bulletin says that in an investigation or market conduct action an insurer can expect questions about how it develops, deploys and uses AI systems, whether or not it has a written AIS Program. Examples it lists include the written program and how it was adopted, the scope of the program including any AI systems it leaves out, model inventories and descriptions, data lineage and bias analysis, validation, testing and model drift evaluation, and, for third-party tools, due diligence, vendor contracts and audits. The bulletin adds that it does not prescribe specific practices or documents. An insurer may show compliance by other means. Keeping these records current and written in plain language is a practical way to prepare.
naic governance relates
How NAIC Governance Relates to SR 26-2 and the NIST AI RMF
SR 26-2 is addressed to banking organizations, not insurers. Its themes can serve as a reference for an insurer's governance program, because the Model Bulletin does not specify a method.
SR 26-2, the interagency model risk management guidance issued in April 2026 that superseded SR 11-7, addresses banking organizations, not insurers. Its themes, model development, validation, and governance, can serve as a reference for an insurer's AI governance program, since the NAIC bulletin leaves the technical method to the carrier.
The NIST AI Risk Management Framework fills the technical layer the bulletin leaves open. Its four functions, govern, map, measure, and manage, give a carrier a concrete way to structure the inventory, risk-tiering, testing, and monitoring work the bulletin calls for in general terms, without requiring the insurer to invent its own taxonomy from scratch.
kriv ai naic
What a Kriv AI NAIC Governance Engagement Includes
A typical engagement starts with the inventory and risk-tiering work, then builds the governance structure, testing protocol, and exam-ready documentation around it, and stays on as an advisory relationship through the carrier's first regulatory exam under the new program.
The starting deliverable is always the same: a complete inventory of AI systems in production or pilot across underwriting, claims, pricing, and marketing, each one risk-tiered and assigned an accountable owner. From there the engagement builds the governance charter and committee structure, a testing and validation protocol matched to each tier, including fairness and proxy-discrimination testing, and the documentation packet a market conduct exam will actually ask to see.
Because the bulletin's language is principles-based, the highest-value part of the engagement is usually translating it into artifacts a compliance team and an examiner can both use: a plain-language system description, a testing summary, and a change log, rather than a purely legal read of the regulation.
engagement
How engagements are scoped and priced
Every engagement is scoped and priced after a short discovery call.
Sources
Cited sources
- NAIC Membership voted to adopt the Model Bulletin on the Use of Artificial Intelligence Systems by Insurers at the 2023 Fall National Meeting, adopted December 4, 2023.
- Full text of the NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers, adopted version.
- NAIC's artificial intelligence insurance topics overview, covering the Model Bulletin and related state regulatory activity.
- NIST AI Risk Management Framework, structured around the Govern, Map, Measure, and Manage functions.
- Federal Reserve SR 11-7, Guidance on Model Risk Management, issued April 4, 2011 and superseded by SR 26-2.
- Federal Reserve SR 26-2, Revised Guidance on Model Risk Management, issued April 17, 2026, superseding SR 11-7.
Related resources
Continue exploring
Straight answers
Frequently asked questions about NAIC AI Model Governance Consulting for Insurers
Who provides NAIC model governance consulting for insurers?
Kriv AI provides AI model governance consulting for insurers, informed by the NAIC bulletin, building the AI system inventory, risk-tiering, testing protocol, and exam-ready documentation a state insurance department's Model Bulletin expects, across underwriting, claims, pricing, and marketing AI systems.
Why is NAIC model governance so hard to implement for AI?
Because the Model Bulletin is principles-based rather than a technical checklist, is adopted state by state with small variations instead of as one national rule, and covers systems spread across actuarial, claims, marketing, and vendor tools that rarely share a common inventory before the governance work starts.
How do insurers implement NAIC model governance for AI?
By building a complete AI system inventory, risk-tiering each system, assigning a named accountable owner and governance committee, defining testing and validation matched to each tier, documenting the program for exam readiness, and monitoring and re-certifying it on an ongoing basis.
What is responsible AI for insurance underwriting?
Responsible AI for underwriting means rating variables are tested for proxy discrimination against protected classes, an adverse decision can be explained in plain language to the consumer and a regulator, a human can review or override higher-risk automated decisions, and each step is documented as it happens.
What is the NAIC Model Bulletin on AI, exactly?
It is the Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, adopted by NAIC membership in December 2023, which sets expectations for how insurers govern AI and predictive models used in consumer-facing decisions and tells state departments what to ask for during an investigation or exam.
Does the NAIC AI Model Bulletin apply in every state?
No. The bulletin is a model document; each state insurance department decides independently whether to issue it and can layer on its own examination guidance, so a multi-state carrier is managing a family of related state-level bulletins rather than one uniform federal rule.
How is NAIC AI governance different from SR 26-2?
SR 26-2, which superseded SR 11-7 in April 2026, is interagency model risk management guidance addressed to banking organizations, not insurers. Insurers can use its development, validation, and governance themes as a reference, because the NAIC bulletin does not specify a technical method.
What is an AIS Program under the NAIC Model Bulletin?
It is the written program the bulletin expects an insurer to develop, implement and maintain for the responsible use of AI systems that make or support decisions in regulated insurance practices. It should address governance, risk management controls and internal audit, and be proportionate to the insurer's use of AI and the potential harm to consumers. It may sit inside the insurer's existing enterprise risk management program.
What can a state insurance department ask for about an insurer's AI?
In an investigation or market conduct action, the bulletin says an insurer can expect questions about its governance framework, risk management and internal controls, and about specific models. Examples include the written AIS Program, model inventories, data and bias analysis, validation and testing records, and, for vendor tools, due diligence and contracts. The bulletin does not prescribe specific documents.
Talk to the team that would do the work
Bring your requirements to a working session with the person who'll actually deliver.
Book a Discovery Call