NAIC’s AI Tool: De-Risking Health Insurance Innovation

Health insurers are pushing artificial intelligence into their core operations fast, and it’s creating a tension between the drive for efficiency and the need to protect consumers. As AI gets baked into everything from claims processing and underwriting to managing patient care, the call for real regulatory oversight is getting louder. This is especially true in healthcare, where a bad algorithm can affect a patient’s health or their ability to pay for treatment. Regulators see what’s happening, and they’re building the tools to actually examine and govern how insurers use AI.

The NAIC’s Proactive Stance on Insurer AI Governance

The National Association of Insurance Commissioners (NAIC) is the main coordinating body for state regulators trying to get a handle on AI. Their work led to the adoption of the Model Bulletin on the Use of AI Systems by Insurers on December 4, 2023. As of the Spring 2026 NAIC meeting, 24 states and the District of Columbia have already adopted this document, which gives them a starting framework for regulating insurance AI. A model bulletin is just paper, though. It needs practical tools for enforcement. That’s where the NAIC’s AI Systems Evaluation Tool comes in.

Developing a Standardized Examination Framework

Throughout 2025, the NAIC’s Big Data and Artificial Intelligence (H) Working Group was busy developing a structured examiner questionnaire called the AI Systems Evaluation Tool. The whole point is to give state insurance regulators a standard way to assess how insurers are governing and validating their AI systems during market conduct examinations, getting everyone on the same page instead of relying on one-off inquiries. The tool is now in a 12-state pilot program (which was expanded to include California) that runs from March to September 2026 with 10 participating insurers to see how it works in the real world. Iowa Insurance Commissioner Doug Ommen, who has been a leader on this, described the tool as optional for regulators but a powerful new resource. This approach respects that states have different regulatory environments while offering a ready-to-use tool. The plan is for the tool to be formally adopted as a standard at the NAIC’s Fall 2026 National Meeting NAIC official meeting schedule and materials.

What the NAIC Evaluation Tool Examines

The AI Systems Evaluation Tool is broken down into four exhibits, giving regulators a 360-degree view of an insurer’s AI practices. These sections get into the weeds of AI governance and deployment, reflecting what’s considered best practice for responsible AI.

Exhibit 1: AI Usage Quantification

First, regulators need a headcount. This exhibit makes insurers list out and quantify their AI usage so regulators have a clear picture of how deeply AI is integrated into their business. Insurers have to identify where AI is being used, what kind of AI it is, and what it’s doing. For a health insurer, this means detailing AI’s role in processing claims, detecting fraud, personalizing health recommendations, or even optimizing its provider network. This inventory helps regulators understand the potential risk attached to an insurer’s AI footprint.

Exhibit 2: Governance Risk Assessment

This is the big one. The governance risk assessment exhibit looks hard at the insurer’s internal policies for managing AI. It examines the company’s governance structure to make sure there are clear lines of responsibility for the AI systems. Regulators will be checking model validation processes, and they’ll want proof that AI models are properly tested for accuracy and reliability before they’re let loose on the public. A huge piece of this is bias testing, which requires insurers to show they are actively finding and fixing algorithmic bias that could lead to discriminatory outcomes or worsen health inequities. The tool also looks at vendor management, since many insurers buy third-party AI and have to make sure their partners meet the same standards.

Exhibit 3: High-Risk AI System Details

This exhibit zooms in on any AI systems the insurer considers “high-risk.” While the definition might change over time, it generally means systems that have a major impact on consumers, like those making decisions about claims approvals, denials, or eligibility for health programs. For these specific systems, regulators expect a lot more documentation covering the AI’s purpose, its design, the data it uses, its performance metrics, and any impact assessments. This information lets regulators really dig in where the potential for consumer harm is greatest.

Exhibit 4: AI Data Details

The final exhibit is all about the data that feeds the AI. Regulators will examine the kinds of data used to train and run the models, as well as the company’s data governance policies. This includes checking data quality, privacy protections (like HIPAA compliance HHS HIPAA compliance guidance), and data lineage. Getting the data details right is non-negotiable, because the quality and representativeness of the data directly control how fair and accurate an AI model is. Poor or biased data can cause models to drift and make existing health disparities even worse.

Moving from Pilot to Standard Practice

The pilot program running from March to September 2026 is the shakedown cruise for the AI Systems Evaluation Tool. Having 12 states and 10 insurers involved provides a diverse test bed, letting the NAIC collect practical feedback and make tweaks. This kind of hands-on development is the only way to create a tool that actually works for regulators without being impossible for insurers to comply with. The expected adoption at the NAIC’s Fall 2026 National Meeting will formalize AI oversight for the industry. And while the tool is technically optional for states, its thoroughness and the NAIC’s backing will likely drive most states to use it. A standard approach helps regulators by giving them a consistent playbook and helps insurers by making expectations clear.

Implications for Buyers, Compliance, and Investors

For people buying health plans, this new regulatory framework means it’s more likely the AI systems affecting their coverage are being checked for fairness and accuracy. It’s a layer of assurance that insurers are being held accountable. If you’re on a health plan’s compliance or governance team, you need to get familiar with these standards now. The NAIC tool is essentially your new internal audit checklist. Insurers that already have strong governance, solid model validation, and good bias testing will be in a much better position when regulators come knocking. This also applies to your vendors, so start doing due diligence on their third-party AI to make sure it’s up to snuff. For investors, the NAIC’s work actually reduces risk by making the regulatory environment more predictable. A standardized examination tool is something you can plan around. Companies that show they take governance seriously, especially around bias testing and consumer impact, will likely be seen as more stable investments less prone to regulatory fines or brand damage. The emphasis on documentation and transparency, as noted in a Fenwick & West analysis of AI regulatory trends, shows the market is maturing. Responsible AI is becoming a regulatory requirement. This work fits in with other things happening, like new state laws on AI and efforts from groups like the AMA to fight AI-driven claims denials. The NAIC tool, however, provides a much broader, system-wide examination of all of an insurer’s AI, laying the groundwork for a health insurance AI market that people can actually trust.