NAIC runs nine‑state AI pilot

The National Association of Insurance Commissioners has moved from writing AI guidance to testing how it works in supervision. In 2026, state regulators are piloting the NAIC’s AI Systems Evaluation Tool to review how insurers use artificial intelligence and whether their governance practices can manage consumer and financial risk. The tool is being used in market conduct exams, financial analysis and financial examinations, according to NAIC materials. The pilot spans property/casualty, life and health insurers and is focused first on domestic carriers. ### So what is the NAIC actually piloting? The NAIC describes the AI Systems Evaluation Tool as a structured set of optional supplemental exhibits for state regulators. The materials say the tool is meant to help regulators identify how extensively a company uses AI, assess governance, drill into higher-risk systems and request model and data details where needed. The exhibits cover AI use across the company, governance risk assessment, high-risk system details and model data details. (content.naic.org) The pilot project summary says the objective is not only to gather information from insurers but also to test whether the tool helps regulators understand AI use and helps insurers explain their governance systems clearly. The same summary says the pilot is intended to support long-term recommendations for market conduct and financial risk assessment review processes. (content.naic.org) ### How many states are in the pilot, and why does that matter? The NAIC’s current public AI topic page says 12 states are participating in the pilot as of March 2026. The pilot project summary available through NAIC materials lists Colorado, Connecticut, Florida, Iowa, Pennsylvania, Rhode Island, Virginia, Vermont and Wisconsin, and says NAIC staff will encourage other interested states to join. (content.naic.org) That gap matters because it shows the program expanded after earlier planning documents were circulated. Fenwick, citing the NAIC program, said California was added and described the pilot as part of a broader effort to shape long-term oversight of insurers’ use of AI and machine learning. (content.naic.org) ### What are insurers being asked to prove? NAIC materials say insurers must be able to demonstrate that risk-based oversight mechanisms are in place and functioning effectively. The tool is designed to supplement existing examination procedures rather than replace them, and regulators are directed to use it alongside existing NAIC handbooks and applicable laws on unfair trade practices, unfair claims settlement, governance, confidentiality and financial reporting. (fenwick.com) The NAIC’s 2023 model bulletin provides the policy backdrop. That bulletin says existing insurance laws apply to AI systems and sets expectations around governance, documentation, testing, transparency and third-party oversight. It also says AI governance should cover uses across the insurance life cycle, including underwriting, pricing, claims administration and fraud detection. (content.naic.org) ### Why is this landing now? The NAIC says its Big Data and Artificial Intelligence Working Group has been developing the evaluation tool in 2025 and 2026 as regulators respond to wider AI adoption in insurance. The group’s 2026 charges include monitoring state, federal and international AI developments and supporting adoption of the model bulletin on insurer AI use. (content.naic.org) A separate pressure point is legal exposure. Insurance Journal reported on May 20 that early litigation is framing some AI disputes not as hacks or ransomware events but as claims arising from ordinary product behavior, such as call analytics, transcription and customer interactions. That argument, as presented in the article, pushes attention toward consent, workflow design, vendor controls and human review rather than cyber-only framing. (content.naic.org) ### Is this just about claims AI? The NAIC says no. Its pilot summary says participating states will include insurers from property/casualty, life and health lines, while the model bulletin says AI oversight should apply across product design, marketing, underwriting, policy servicing, claims and fraud detection. Claims Journal, in a sponsored article published May 20, said the pilot reflects a broader shift in insurance AI discussions toward fairness, accuracy, accountability and auditability in claims operations. (insurancejournal.com) That article is not an NAIC document, but it aligns with the regulator’s published emphasis on governance and explainability. The next public marker is the 2026 Fall National Meeting. (content.naic.org) The NAIC’s AI topic page says the tool is anticipated to be adopted then, after pilot feedback is reviewed, and the pilot summary says regulators will provide updates at each National Meeting. (content.naic.org) (claimsjournal.com)