NAIC Creates Market Conduct Working Group to Examine AI-Era Insurers

From monitoring NAIC committee rosters, working group charges, and meeting agendas across quarterly meetings, the formation of the Market Conduct Regulation Modernization (D) Working Group at the Spring 2026 National Meeting signals that regulators now recognize the exam framework itself, not just the rules being examined, needs updating. The Market Regulation and Consumer Affairs (D) Committee appointed the group on March 25, 2026 in San Diego, with a mandate to deliver actionable recommendations by the Fall 2026 National Meeting. This is the first NAIC body explicitly tasked with rethinking how state regulators examine insurers in an industry where algorithmic pricing, automated claims handling, AI-generated customer communications, and technology-enabled distribution channels no longer fit the legacy review frameworks that most state market conduct exam teams still use.

Why a New Working Group, and Why Now

Market conduct examination has not received a structural overhaul comparable to what solvency regulation got after the 2008 financial crisis. The Solvency Modernization Initiative produced the ORSA requirement, group supervision frameworks, and risk-based capital reforms. The market conduct side, by contrast, has relied on incremental updates to the Market Regulation Handbook and the Market Conduct Annual Statement (MCAS) data collection while the industry it regulates underwent a fundamental technology transformation.

Illinois Department of Insurance Director Ann Gillespie, who chairs the parent Market Regulation and Consumer Affairs (D) Committee, framed the initiative plainly: "State insurance regulators focus on leading, not just responding. That means undertaking a comprehensive assessment of the current market regulatory framework and identifying opportunities to enhance policyholder protection, while creating efficiencies for regulators and companies." That language was carefully chosen. The dual emphasis on consumer protection and regulatory efficiency reflects a core tension in market conduct modernization: expanding oversight to cover AI-driven processes while streamlining exam procedures that insurers have long complained are duplicative and inconsistent across states.

NAIC President and Virginia Insurance Commissioner Scott A. White reinforced this framing in his Spring 2026 keynote address. White described the initiative as developing "a strategic, forward-thinking roadmap to determine if today's market conduct regulatory framework has the data, the system, the tools, and the supervisory approaches needed to oversee a rapidly evolving industry." He noted that artificial intelligence, new distribution models, national-scale vendors, and other technological advances are "significantly changing both consumers' expectations and insurers' business models and practices."

Five Mandated Work Areas

The working group's charge covers five specific domains. Each addresses a structural gap that predates the current AI acceleration but that AI has widened considerably.

Data collection and analysis. The MCAS, which standardizes the data states use to benchmark insurer market behavior, was designed for a world of human-processed applications and manual claims review. When a carrier deploys a gradient-boosted model to triage auto claims or uses a large language model to draft policyholder correspondence, the MCAS data fields do not capture the operational shift. The working group will evaluate whether current data collection captures enough signal about AI-driven processes to allow regulators to identify patterns that warrant a full exam. The NAIC's 2026 strategic priorities already flag "enhancing data architecture, predictive analysis, and market analysis" as a top institutional goal, and the modernization working group is the vehicle for translating that aspiration into specific data reporting changes on the market conduct side.

Interstate collaboration. A multi-state carrier may face market conduct exams in a dozen jurisdictions, each applying slightly different standards, requesting slightly different documentation, and arriving at slightly different conclusions about identical business practices. The working group has explicit authority to address coordination gaps in both data analysis and examination. For carriers running centralized AI systems, where a single underwriting model serves policyholders across 40 states, inconsistent examination standards create a structural problem: a model deemed compliant in one exam may be flagged in another, with no mechanism to reconcile the findings. The working group's charge to improve interstate collaboration is arguably the single most impactful element for multistate insurers.

Market Regulation Handbook and examination approaches. The Handbook is the primary reference for exam teams conducting market conduct reviews. It was last updated to address emerging technologies in limited ways: the Market Conduct Examination Guidelines (D) Working Group has been drafting examiner guidance for pet insurance and incorporating recommendations from the disbanded Accelerated Underwriting (A) Working Group. The modernization working group's scope is far broader. It must evaluate whether the Handbook's examination methodologies, designed for paper-file reviews and claims sampling, can adequately assess algorithmic decision-making, automated customer interactions, and AI-driven underwriting at scale.

NAIC systems and support. The technology infrastructure that state regulators use to conduct market analysis and coordinate exams has not kept pace with the tools insurers deploy. The working group will assess whether NAIC systems provide sufficient analytical capability for regulators to monitor the outputs of AI-driven insurance operations. This is a capacity question as much as a technology question: the SupTech/GovTech Subgroup under the Innovation, Cybersecurity, and Technology (H) Committee has been hearing presentations from state data and analytics teams, but those efforts have focused on individual state capabilities rather than a coordinated platform upgrade.

Examiner training. Market conduct examiners historically needed deep knowledge of insurance law, claims handling procedures, and underwriting standards. Examining AI-driven insurers adds a layer: examiners must now understand model governance, algorithmic fairness testing, data lineage, and third-party vendor oversight, at least well enough to ask the right questions. The working group's training mandate connects directly to the NAIC's 2026 strategic priority of "expanding capacity and expertise through education and training." The 12-state AI evaluation pilot already surfaced training needs, as pilot states received specialized instruction on using the AI Systems Evaluation Tool, and the modernization group will need to institutionalize those lessons into the standing examiner training curriculum.

How This Differs From Existing NAIC AI Bodies

The NAIC has multiple committees and working groups touching AI governance. Understanding where the Market Conduct Regulation Modernization Working Group fits requires mapping the organizational landscape.

The Big Data and Artificial Intelligence (H) Working Group sits under the Innovation, Cybersecurity, and Technology (H) Committee. This is the body that developed the AI Systems Evaluation Tool, manages the 12-state pilot running from March to September 2026, proposed the four-tier AI risk taxonomy (unacceptable, high, medium, low), and is advancing the operationalization of the December 2023 Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, now adopted by 23 states plus Washington, D.C. The BDAI Working Group focuses on the what: what standards should govern insurer AI, what information regulators need to assess compliance, and what enforcement mechanisms should apply.

The Innovation, Cybersecurity, and Technology (H) Committee itself provides broader technology oversight, spanning cybersecurity (including the new Centralized Cybersecurity Event Notification Portal), data privacy, and technology adoption by both insurers and regulators.

The Third-Party Data and Models (H) Working Group, also under the (H) Committee, is developing a vendor registry framework. In Phase One, the group narrowed its focus from six insurance functions (pricing, underwriting, claims, utilization reviews, marketing, fraud detection) to pricing and underwriting specifically. The outstanding questions, whether registration should be mandatory or voluntary, and whether the framework should be guidance or model law, remain unresolved.

The Market Conduct Regulation Modernization Working Group operates in a different lane. It sits under the Market Regulation and Consumer Affairs (D) Committee, not the (H) Committee. Its focus is not the substance of AI rules but the process by which regulators examine whether insurers follow those rules. The distinction matters practically: the BDAI Working Group could produce a comprehensive AI governance framework, and the Third-Party Working Group could build a complete vendor registry, and market conduct examiners would still lack the tools, training, data collection standards, and coordination mechanisms to verify compliance during actual examinations. The modernization group fills that gap.

The relationship is sequential. The BDAI Working Group defines what carriers must do with AI. The modernization group defines how regulators check whether carriers are doing it. That second question has not had a dedicated institutional home until now.

The 12-State AI Pilot: Input Source, Not Answer

The AI Systems Evaluation Tool pilot, running across California, Colorado, Connecticut, Florida, Iowa, Louisiana, Maryland, Pennsylvania, Rhode Island, Vermont, Virginia, and Wisconsin from March through September 2026, generates real-world data about how examiners interact with insurer AI documentation. Pilot states meet monthly to share progress, and participating regulators coordinate to prevent duplicative requests to the same carrier.

The pilot's four-exhibit structure (Exhibit A on AI system inventory, Exhibit B on governance frameworks, Exhibit C on high-risk model details, and Exhibit D on data inputs and lineage) provides a structured examiner questionnaire. But it does not address the structural questions the modernization group must tackle. The pilot asks: can this tool extract useful information from an insurer? The modernization group asks: should market conduct exams even be structured the way they are when the insurer being examined runs fundamentally different operations than the legacy model the exam framework assumed?

Consider the timing. The pilot is expected to produce results in September through October 2026, with tool revisions and public re-exposure before adoption at the Fall 2026 National Meeting. The modernization group must deliver its initial recommendations at the same meeting. The two streams will converge: pilot findings about where the Evaluation Tool works and where it falls short will inform the modernization group's recommendations about exam methodology, data collection, and examiner training. The pilot is an input to the modernization process, not a substitute for it.

What Specific Exam Changes Could Emerge

The working group has a broad mandate and a tight timeline. Based on the charges, the institutional context, and the gaps that recent NAIC activity has surfaced, several categories of recommendations are plausible.

Standardized AI-related MCAS data fields. The most immediately actionable recommendation would be expanding the Market Conduct Annual Statement to include data fields capturing the scope of AI deployment, model governance incidents, consumer complaints tied to automated decisions, and third-party vendor usage. These fields would give regulators a systematic data source for identifying which carriers warrant closer examination, rather than relying on consumer complaints or media coverage to trigger an AI-focused exam.

Model governance documentation standards for exam readiness. The current Market Regulation Handbook does not specify what documentation examiners should request when evaluating an AI-driven process. Carriers that have already responded to the AI Evaluation Tool pilot know the four-exhibit structure, but that tool was designed for a specific pilot context. The modernization group could establish permanent documentation standards, specifying what model cards, validation records, bias testing results, and drift monitoring reports examiners should routinely request during any market conduct exam involving automated decision-making.

Coordinated multi-state examination protocols. For carriers operating centralized AI systems across multiple states, the working group could propose protocols enabling lead-state examinations with multi-state recognition, similar in concept to financial exam coordination under the NAIC accreditation program. A carrier whose underwriting model is examined by its domiciliary state under agreed-upon standards would not face redundant examinations of the same model in every other state where it writes business. This would require states to agree on minimum examination standards and accept the results of other states' work, a politically difficult proposition but one that the explicit interstate collaboration charge puts on the table.

Examiner competency requirements for AI-related exams. The training mandate could produce specific competency standards that examiners must meet before participating in market conduct examinations of AI-driven insurers. The pilot has already demonstrated that examining AI governance requires different skills than examining claims files. The American Academy of Actuaries, in its comment letter on the AI Model Law RFI, emphasized the importance of technical competency in regulatory oversight. The modernization group could establish certification or continuing education requirements for examiners conducting technology-focused reviews.

Trigger-based exam scheduling tied to AI deployment metrics. Traditional market conduct exam scheduling relies on a combination of MCAS data analysis, consumer complaint trends, and periodic review cycles. The working group could recommend risk-based triggers specifically tied to AI deployment: a carrier that introduces a new automated claims system, onboards a third-party AI vendor for underwriting, or significantly expands the scope of algorithmic decision-making could face expedited review. This would align exam frequency with the pace of technology change rather than relying on lagging indicators.

The Interstate Coordination Problem in Detail

Interstate coordination is where the modernization group's potential impact is highest and the political difficulty greatest. Consider the current landscape: 23 states plus D.C. have adopted the NAIC's AI Model Bulletin, but adoption varies in scope and specificity. Colorado's AI Act (SB 21-169 and the more recent SB 26-189) imposes quantitative bias testing requirements that exceed most other states' mandates. Connecticut's SB 5 creates a comprehensive AI governance framework. Texas passed TRAIGA with its own disclosure requirements. The 12-state evaluation pilot adds a fourth compliance layer.

A multi-state insurer using one predictive underwriting model across all jurisdictions faces at least four distinct compliance regimes. Market conduct examiners in each state may evaluate the same model against different standards. Without coordination, the insurer produces different documentation packages for each exam, examiners duplicate each other's work, and inconsistent findings create legal and operational uncertainty.

The modernization group's interstate collaboration charge is an acknowledgment that this fragmentation is unsustainable, particularly as AI systems become more central to insurer operations. But the solution requires more than a technical fix. It requires states to agree on what constitutes an adequate examination of an AI system and to accept other states' work as satisfying their own examination requirements. The NAIC accreditation program provides a precedent on the financial examination side, but market conduct has historically resisted the same degree of standardization because states view consumer protection as a core sovereignty prerogative.

The Fall 2026 deadline creates a forcing function. The working group does not need to solve the full coordination problem by November, but it can propose a framework, perhaps a model coordination agreement or a set of minimum examination standards, that states can adopt voluntarily. The 12-state pilot already demonstrates that multi-state coordination is possible when regulators choose to participate. The question is whether the modernization group can institutionalize that coordination beyond a pilot context.

What This Means for Carriers and Actuarial Teams

The Fall 2026 recommendation deadline creates a concrete planning window. Carriers and their actuarial teams should be acting now on several fronts.

Audit your model inventory. The pilot's Exhibit A exposed a widespread gap: most carriers do not maintain a comprehensive, defensible inventory of all AI and ML models in production. If the modernization group recommends standardized MCAS data fields capturing AI deployment scope, carriers will need to report model counts, deployment contexts, and consumer-facing applications in a structured format. Building that inventory now, before it becomes a regulatory requirement, is the single highest-return compliance investment.

Consolidate examination documentation. If interstate coordination recommendations emerge, carriers that maintain separate documentation packages for each state's examination requirements will need to consolidate into a single, comprehensive governance file that satisfies the most rigorous state standard. Actuarial teams involved in model validation should be documenting bias testing results, drift monitoring procedures, and third-party vendor oversight in a format designed for examiner consumption, not just internal risk management.

Engage in the stakeholder input process. Director Gillespie emphasized that the NAIC will take "an inclusive and deliberate approach to collect broad input throughout 2026." Carriers, trade groups, and individual actuaries have an opportunity to shape the recommendations before they crystallize. The modernization group's charges are broad enough that stakeholder input will significantly influence how those charges translate into specific recommendations. Companies that wait for final output before responding will have less influence than those that participate in the development process.

Track the pilot convergence. Because the AI Evaluation Tool pilot and the modernization group both target the Fall 2026 National Meeting for their deliverables, the two streams will interact. Pilot findings about examiner capacity, tool effectiveness, and documentation gaps will directly inform the modernization group's work. Carriers participating in the pilot should be tracking not just their own examination experience but the broader pilot lessons that will shape exam methodology recommendations.

Prepare for examiner competency changes. If the modernization group establishes new training requirements for examiners conducting AI-focused reviews, the sophistication of examination questions will increase. Actuarial teams that currently interact with market conduct examiners should anticipate questions about model governance that go beyond the current Handbook's scope: algorithm selection rationale, feature engineering choices, fairness metric selection, and the actuarial judgment behind threshold settings in automated decision systems.

The Broader Regulatory Trajectory

Patterns we have tracked across quarterly NAIC meetings show a clear institutional acceleration. The December 2023 Model Bulletin established principles. The May 2025 Request for Information tested appetite for enforceable rules. The 12-state pilot launched in March 2026 to operationalize the Evaluation Tool. The Third-Party Data and Models Working Group is building vendor oversight. And now the Market Conduct Modernization Working Group targets the examination framework itself.

Each step builds on the prior one, and each addresses a different layer of the regulatory stack. The Model Bulletin says what carriers should do. The potential model law (still under deliberation) would say what they must do. The Evaluation Tool gives examiners a questionnaire for checking compliance. The vendor registry would extend oversight to the AI supply chain. The modernization group now asks whether the exam framework, the mechanism through which all of those requirements actually get enforced, is adequate for the task.

This is the regulatory equivalent of discovering that you have a building code, an inspection checklist, and trained inspectors, but the inspection schedule was designed for wood-frame houses and you are now inspecting prefabricated steel structures. The building code might be sound, the checklist might cover the right items, and the inspectors might know what to look for. But the process of getting inspectors into the building, coordinating across jurisdictions, and collecting the data needed to identify problems, that process was built for a different type of construction.

The Fall 2026 recommendations will not be the final word. The working group is expected to provide "initial" recommendations, suggesting that further work will continue into 2027 and beyond. But the initial recommendations will set the direction: whether the NAIC moves toward coordinated multi-state AI examinations, whether MCAS data collection expands to capture AI-specific metrics, and whether examiner training evolves to match the technology sophistication of the insurers being examined. Those directional choices will shape compliance investments across the industry for years.