NAIC Embeds Life AI Exam Guidance Into Market Conduct Reviews

Fifty-nine percent of individual life insurance applications now route through accelerated underwriting programs, bypassing the blood draw and attending physician statement that defined new business issuance for a generation. The Life Insurance and Annuities (A) Committee adopted regulatory guidance on August 14, 2024 that specifically targets the predictive models, external data sources, and no-exam decision algorithms approving those applications, and that guidance is now being embedded into the Market Regulation Handbook that state examiners will carry into 2026 market conduct reviews. The timing marks a threshold: 2026 is the first full examination cycle in which life actuaries and their AUW program governance teams can expect to face structured scrutiny of the algorithms that now underwrite the majority of term and simplified whole life business.

From reviewing accelerated underwriting program governance documentation at life carriers preparing for market conduct exams in 2025, the gap between what the algorithm does and what the actuarial documentation explains is often 12 to 18 months wide. A carrier deploys a revised mortality scoring model, tightens its prescription database lookup logic, or adjusts its MVR-based hazard classification over a product development cycle that moves faster than the actuarial opinion cycle. The August 2024 NAIC guidance was written specifically to close that gap, and the 2026 exam cycle is the first time state examiners will have a formal framework for measuring it. That combination, a structured examination methodology applied to a technology that has been running ahead of its documentation, is why the guidance is generating more carrier attention in mid-2026 than it did when it was first adopted nearly two years ago.

The August 2024 Adoption and What Followed

The Accelerated Underwriting (A) Working Group spent several years developing regulatory guidance before bringing it to a vote. The Life Insurance and Annuities (A) Committee adopted the final guidance package on August 14, 2024, at the NAIC Summer National Meeting, simultaneously adopting a referral that directed the Market Conduct Examination Guidelines (D) Working Group to update the Market Regulation Handbook to incorporate the AUW framework. The working group was disbanded shortly after, its mission complete. What remained was the integration work, and that work is now the operative fact for 2026 examinations.

The Market Conduct Examination Guidelines (D) Working Group carries ongoing authority over the Market Regulation Handbook, the primary reference document state market conduct examiners use during insurer reviews. Embedding the August 2024 AUW guidance into that handbook converts regulatory principles into examination procedure. It tells examiners not just that accelerated underwriting programs should comply with certain standards, but what to request, how to evaluate the responses, and where to escalate. That procedural detail is what life actuaries should be reading, and many have not yet done so.

The NAIC Spring 2026 National Meeting in San Diego (March 22-25) added a further layer. A new working group formed at that meeting is expected to deliver structural examination methodology recommendations to the full (D) Committee for adoption at the Fall 2026 National Meeting in November. That means the examination framework is still being refined, and carriers with AUW programs operating in states that move quickly to implement the Fall 2026 recommendations face a compliance landscape that will tighten further in 2027.

Three Areas of Focus the Guidance Defines

The regulatory guidance package the Life Insurance and Annuities Committee adopted organizes its requirements around three areas of focus: data sources used as exam substitutes, predictive models and their validation, and the no-exam decision logic itself. Understanding each area separately is the right starting point for a life actuary building a compliance program, because each area maps to a different set of internal documentation obligations and a different pattern of vendor dependency.

The data source requirements address the credit scores, prescription history databases, motor vehicle records, and consumer data aggregator files that AUW programs use to replicate the underwriting information a paramedical exam would have produced. The guidance requires that these sources be clearly identified, that the carrier be able to explain the actuarial basis for using each source as an exam substitute, and that the carrier have a mechanism for consumers to challenge data points that appear incorrect or that influenced an adverse underwriting decision. Carriers running AUW programs that use five or six external data inputs but have never produced a consolidated inventory of those inputs and their documented rationale are in the position the guidance was designed to correct.

The predictive model requirements address the mortality scoring models, risk classification algorithms, and automated approval engines that translate raw data inputs into an underwriting decision. The guidance requires validation procedures and discrimination testing as explicit deliverables. Validation in this context means demonstrating that the model produces the mortality outcomes it was designed to predict, not just at initial deployment but on an ongoing basis as the in-force book develops. Discrimination testing means evaluating whether the model produces outcomes that are unfairly discriminatory across protected classes, including an analysis of whether variables that appear facially neutral function as proxies for race, national origin, or sex. That second requirement has no direct analog in traditional life underwriting actuarial work, and many AUW programs entered 2026 without a documented discrimination testing protocol.

The no-exam decision logic requirements address the approval algorithm itself, the rules and model outputs that determine whether an application receives an automated approval, a request for additional information, or a referral to full traditional underwriting. The guidance requires documentation of how that logic operates, what overrides exist, and how exceptions are handled. The practical implication for life actuaries is significant: many AUW approval engines were built iteratively, with rules layers added over product development cycles, and the full logic tree is often held partly in model code and partly in institutional memory rather than in a single documented specification.

The Volume That Makes the Guidance's Reach Enormous

Gen Re's 2024 U.S. Individual Life Next Gen Underwriting Survey, drawing on 30 individual life carriers representing more than 2 million paid policies and $827 billion in coverage volume, shows how thoroughly accelerated underwriting has displaced the paramedical exam as the default issuance path. An average of 59% of individual life insurance applications qualify for an accelerated underwriting path, with the range running from 15% to 100% across participating carriers. Eighty-two percent of carriers have either a fully or partially implemented AUW workflow, and 10% more plan to implement within two years. The traditional fully-underwritten new business process is now the exception for individual term life, not the rule.

The efficiency gain that drove adoption is real and durable. Carriers in the Gen Re survey average a five-day underwriting decision under AUW compared to a 23-day decision under full traditional underwriting, an 18-day improvement per application. That speed advantage translates directly into placement rates and producer satisfaction metrics that carriers have built into their distribution relationships. The business case for AUW is settled. The regulatory question for 2026 is whether the documentation supporting those programs meets the standard the August 2024 guidance establishes.

Ninety-four percent of survey participants offer term life through AUW channels, and 63% offer universal life and whole life as well. Sixty-three percent of carriers use random holdouts as statistical controls to monitor whether AUW-approved applicants develop mortality experience consistent with what full underwriting would have produced. Ninety-one percent currently use or plan pre-issue and post-issue auditing as control mechanisms. Those control statistics are relevant to the examination framework because they describe the monitoring infrastructure the guidance requires carriers to have in place. A carrier that uses holdouts and conducts post-issue audits has an empirical basis for defending its AUW model's ongoing validity. A carrier that deploys AUW at scale without systematic mortality slippage monitoring has a gap the examiner's documentation request will expose.

Carrier Accountability When the Algorithm Belongs to a Vendor

The accountability question is the sharpest edge of the August 2024 guidance, and also the one that generates the most internal confusion at carriers with mature AUW programs. When a life carrier uses an outside vendor's mortality scoring model as the basis for an AUW approval, the guidance places regulatory responsibility for that model on the carrier, not the vendor. The NAIC Model Bulletin adopted in December 2023, which is now in effect in more than 25 states, establishes the same principle for AI and predictive model use generally. The carrier that deploys the model to make or materially influence an underwriting decision carries the documentation and compliance obligation, regardless of whether the underlying algorithm was built in-house or licensed from a third-party data and analytics provider.

The practical consequence is that life actuaries at carriers using vendor-supplied mortality scoring products from providers such as LexisNexis Risk Solutions or MIB Group need to be able to produce documentation they may not currently have. An examiner asking for the validation methodology and discrimination testing results for a vendor model is asking the carrier to demonstrate oversight of something the vendor controls. Standard vendor agreements for insurance data analytics products do not typically include provisions that give the insurer access to detailed model validation documentation, training data composition, or proxy discrimination analysis. The carrier signs a data license and receives an API endpoint; the actuarial rigor of what runs behind that endpoint is treated as the vendor's proprietary asset.

That contract structure is incompatible with what the August 2024 guidance requires in an examination context. Carriers that receive market conduct exam requests in 2026 and discover that their largest AUW data vendor will not provide the model documentation an examiner needs are facing a gap that was foreseeable when the guidance was adopted. The practical response, for carriers still in contract renewal cycles, is to negotiate audit rights that allow the carrier to commission or receive independent validation analysis for any vendor model that drives underwriting decisions at the scale the Gen Re survey describes. The NAIC's Third-Party AI Vendor Registry framework, which is being developed in parallel to the examination guidance integration, is designed to create a centralized documentation layer for vendor AI and predictive models over time. That registry does not yet exist in a form carriers can rely on for 2026 examination responses. The contractual work is what bridges the gap now.

State-by-State Adoption: Where the Compliance Landscape Is Uneven

The August 2024 NAIC guidance carries the authority of an NAIC adoption, but the Market Regulation Handbook is a best-practices reference that each state implements through its own examination protocols. Not every state has updated its market conduct examination procedures to reflect the August 2024 AUW referral with the same speed. Some states moved quickly to embed the guidance into their exam procedures for life insurer reviews beginning in 2025. Others have not yet updated their examiner training and request templates to reflect the AUW-specific documentation requirements. A carrier operating in 30 or 40 states may find that the compliance exposure varies significantly depending on which state's examiners are active in a given review cycle.

The states with the most active market conduct examination programs for life insurance, including California, New York, Illinois, and Texas, each maintain their own examination frameworks and have their own timelines for integrating NAIC handbook updates. New York, which operates under its own Insurance Law and does not always adopt NAIC model guidance on the same schedule as other states, maintains examination authority over life insurer AUW programs that has been exercised independently of the NAIC's working group timeline. Carriers with large New York life business should not assume that New York examination standards track directly to the August 2024 guidance timeline that applies elsewhere.

The Spring 2026 NAIC meeting's new working group adds another variable. If that group delivers substantive recommendations for Fall 2026 adoption, states that move quickly to implement those recommendations will begin applying a revised, and likely more detailed, examination standard in early 2027. Carriers that build their 2026 compliance documentation to the current guidance may find themselves preparing a second documentation round within 12 months. The carriers best positioned for that scenario are those that document not just to the current guidance standard but to the principle behind it: that every component of an AUW program should be explainable to a technically sophisticated examiner in a single session, without requiring weeks of internal discovery to locate the underlying analysis.

The Documentation Gap: 18 Months Between Model and Memo

The most consistent finding from life carrier AUW compliance preparedness reviews is the temporal gap between what the AUW algorithm does in production and what the actuarial documentation explains. A life carrier launches an AUW program with an initial mortality scoring model and supporting actuarial analysis. Over the next 18 months, the product development and data science teams iterate on the model, refining prescription database lookup rules, adding a new MVR severity weighting, adjusting the credit-based score threshold, and recalibrating the automated approval boundary. Each change is tested for mortality impact and approved through an internal model governance process. The actuarial opinion that originally described the model is not updated with each change; it gets refreshed on a periodic cycle, often tied to product pricing reviews rather than to model change events.

The result, which state examiners applying the August 2024 guidance will encounter repeatedly in 2026, is that the most current version of the AUW approval engine is materially different from what the most current actuarial documentation describes. The examiner asks: what model is producing these no-exam approval decisions? The actuarial team produces a document that accurately described the model as it existed 18 months ago. The gap between the document and the live system is not evidence of bad faith; it reflects a documentation cycle that was designed for product pricing reviews, not for a regulatory examination framework that treats each model update as a potentially material change requiring documentation refresh.

ASOP No. 56 (Modeling), which governs what an actuary must do when developing, selecting, modifying, or reviewing a model, provides the applicable professional standard. The standard requires actuaries to assess the model's reasonableness and appropriateness for its intended purpose, evaluate input data quality, and disclose material limitations. In an AUW context, each material modification to the approval algorithm, a recalibrated mortality score threshold, a revised prescription database weighting, a new external data source substituting for a prior one, is a model change that ASOP 56 documentation practices should capture. Carriers that treat ASOP 56 documentation as a point-in-time opinion rather than an ongoing obligation for a production scoring system are creating the documentation gap that the August 2024 guidance specifically targets.

The practical response is to establish a model change log for each AUW component that is updated when the model changes, not only when the actuarial opinion is refreshed. That log does not need to be a full actuarial opinion for each change; it needs to be a structured record that identifies what changed, when it changed, what testing supported the change, and whether the change was material enough to require a full documentation refresh. Carried forward consistently, that log gives an examiner a traceable history of every modification to the live AUW system, with supporting analysis for each change. It converts a documentation gap into a documentation trail.

What the Examination Request Will Actually Ask For

Life actuaries whose carriers receive a market conduct examination request that invokes the August 2024 AUW guidance should expect requests organized around the guidance's three areas. Data source requests will ask for an inventory of every external database and data feed used as an underwriting input, the documented actuarial basis for each source, the consumer challenge and correction mechanism, and evidence that the carrier has tested each source for accuracy and representativeness in its applicant population. For carriers using prescription history from a major pharmacy benefit database, MIB coding, credit-based scores, motor vehicle records, and one or more geospatial or consumer data enrichment feeds, this inventory will be substantial, and assembling it for the first time under examination pressure is a multi-week project.

Predictive model requests will ask for the validation methodology for each model in production use in the AUW program, the discrimination testing results, and a description of how the carrier monitors for ongoing compliance with the unfair discrimination standard. The validation request is directed at both internally developed models and vendor-supplied models. For internally developed models, the carrier typically has this documentation in some form, distributed across actuarial and data science team files. For vendor-supplied models, the carrier will need to demonstrate that it has exercised meaningful oversight of the model's statistical validity and discrimination outcomes, not simply licensed and deployed a vendor product without independent review.

Decision logic requests will ask for a description of the approval algorithm: what model outputs and decision rules determine whether an application is auto-approved, auto-declined, referred for additional requirements, or routed to full underwriting. The request will typically also ask for documentation of human override capabilities and the frequency with which overrides occur, the adverse action notification procedures for applicants who receive a worse underwriting classification or decline based on AUW model outputs, and the monitoring records that demonstrate the algorithm produces outcomes consistent with the actuarial assumptions supporting product pricing. For a carrier with a complex AUW engine built incrementally over several product development cycles, producing a single coherent description of the decision logic requires the kind of cross-functional documentation review that rarely happens outside of examination preparation.

The NAIC's 12-state AI Systems Evaluation Tool pilot, running from March through September 2026, is generating parallel documentation requests for AI and predictive models across all business functions, including life underwriting. Carriers already engaged in the pilot have had direct experience with what examiners ask for when they have a structured AI examination framework. The AUW guidance examination integration is a life-specific analog to the same dynamic: a structured documentation request framework being applied to a business function that has been running AI-enabled decision systems without that examination layer in place.

Connecting the NCOIL Genetic Testing Parallel

The August 2024 AUW guidance does not operate in isolation from other regulatory developments that affect life underwriting data and decision logic. The NCOIL Life Insurance Genetic Testing Model Act, heading toward a Spring 2026 vote, defines "genetic test" narrowly in a way that preserves accelerated underwriting pipelines that use Rx data and MVR records rather than direct genetic testing results. That definitional clarity matters for AUW compliance programs: carriers can document that their prescription database lookups, MIB coding, and credit-based score inputs fall outside the genetic testing definition, and therefore outside the NCOIL model's antiselection restrictions, without reconstructing the AUW data source inventory that the August 2024 guidance requires separately.

The overlap point is documentation: a life actuary building a complete AUW compliance file for the August 2024 guidance exam requirements will produce much of the same inventory that a NCOIL genetic testing compliance analysis needs. Mapping each AUW data input to its actuarial basis and confirming that none of the inputs constitute a "genetic test" under the model act's definition is additive work on top of the August 2024 documentation, not a separate project. Carriers that treat AUW compliance documentation as a single integrated project, rather than as a series of separate regulatory responses, will produce a more complete and defensible file and avoid duplicating the inventory and testing work across multiple compliance workstreams.

The Actuarial Function's Ownership of the Documentation Obligation

One consequence of how AUW programs have developed at many carriers is ambiguity about which function owns the regulatory compliance obligation. Data science teams built the mortality scoring models. Product development teams defined the approval boundaries. Technology teams built the integration with external data vendors. Legal and compliance teams monitor regulatory guidance. The actuarial function produced the initial pricing and reserving analysis and, in some cases, filed rate and form documentation with state regulators. The August 2024 guidance sits at the intersection of all of those functions, and the examination request, when it comes, will arrive at the legal or compliance department before reaching the actuarial team.

The NAIC Model Bulletin's framework, which the AUW guidance is aligned with, places primary accountability on the insurer's "responsible AI governance" program, which typically sits within legal, compliance, or an enterprise risk function. But the documentation the guidance specifically requires, model validation, discrimination testing, mortality assumption support, data source actuarial basis, is documentation that only the actuarial function can credibly produce. Life actuaries who have defined their role in the AUW program narrowly, as pricing advisors or reserving technicians rather than as ongoing model owners, may find that the examination request arrives at their desk indirectly, filtered through a compliance function that does not itself have the technical background to assemble the response.

The cleaner structure, which the August 2024 guidance implicitly requires, is for the appointed actuary or chief actuary to hold explicit ownership of the AUW model documentation obligation: not just the initial pricing analysis, but the ongoing model change log, the periodic validation refresh, the discrimination testing program, and the data source inventory with actuarial rationale. That ownership claim is consistent with ASOP 56's scope and with the professional responsibility framework that applies to actuaries who rely on predictive models in their work. It also clarifies who speaks for the AUW program when a market conduct examiner asks.