The National Association of Insurance Commissioners is broadening a pilot program that evaluates how insurers use artificial intelligence. The expansion comes as regulators face increased pressure to understand and govern predictive models now embedded in underwriting, claims, and pricing decisions across the industry.
The pilot, initially focused on a narrower set of AI applications, will now include machine learning systems - a subset of AI that learns from data without explicit programming. This shift reflects the reality that most modern insurance AI tools rely on machine learning techniques rather than static rule-based algorithms.
State insurance regulators have been wrestling with how to ensure these systems comply with existing market conduct and unfair trade practices laws. Unlike traditional actuarial models, machine learning algorithms can surface correlations that may inadvertently act as proxies for protected characteristics.
The scope of the expanded pilot
The NAIC's move signals an intent to build regulatory capacity around technical evaluation. By including machine learning, the pilot acknowledges that the complexity of these models requires new testing frameworks. Regulators are not simply checking for compliance with written policies - they are examining the actual behavior of deployed algorithms.
Insurers participating in the pilot will submit their models for review. The NAIC has not disclosed which carriers are involved, but the association's membership includes commissioners from all 50 states, the District of Columbia, and five U.S. territories.
Industry context and regulatory momentum
The expansion arrives amid broader scrutiny of AI in financial services. Federal agencies and state regulators alike have issued guidance on algorithmic fairness, and the insurance sector sits at the intersection of consumer protection and risk-based pricing. Insurers rely on predictive models to price policies, detect fraud, and triage claims - areas where bias can produce real harm.
For professionals working in AI for Insurance, the pilot represents a regulatory signal worth watching. The frameworks developed through this process could influence model governance requirements across multiple lines of business.
State-level action often moves faster than federal legislation, and insurance regulation has historically been state-led. The NAIC's work may set de facto standards that carriers adopt nationwide.
Why this matters for insurance professionals
Compliance teams, actuaries, and data science units should expect model documentation requirements to grow more granular. If the pilot yields a formal evaluation framework, carriers will need to demonstrate not just what a model predicts, but how it arrives at those predictions - and whether the inputs introduce unlawful discrimination. Early preparation for technical audits will reduce friction when regulators begin formal reviews.
Your membership also unlocks:
