AI in Health Insurance Scrutiny

States are moving to police how health insurers use artificial intelligence to approve, delay or deny care. California already has a law on the books, and Pennsylvania and New Hampshire are weighing new rules now. (natlawreview.com) (legiscan.com) The bills target prior authorization and utilization review, the insurer processes that decide whether a treatment is medically necessary before or after care is delivered. In Pennsylvania, House Bill 1925 and Senate Bill 1113 would let insurers use artificial intelligence, but not to override a clinician’s judgment. (natlawreview.com) Those Pennsylvania bills would also require a healthcare provider to review the patient’s records before any adverse benefit determination, and would force insurers to file annual artificial intelligence compliance statements and disclose artificial intelligence use to members and providers. The Senate version adds a penalty that can temporarily block new enrollments for noncompliant insurers. (natlawreview.com) New Hampshire’s House Bill 1406 would treat artificial intelligence as an assistive tool, not the final decision-maker. It would require written records on how the tool was used and says a qualified healthcare provider must make adverse determinations. (natlawreview.com) California moved first. Senate Bill 1120 was approved on September 28, 2024, and requires health plans and disability insurers using an artificial intelligence, algorithm or software tool in utilization review to ensure the tool relies on specified information and is applied fairly and equitably. (legiscan.com) The state push sits on top of a broader insurance rulebook that regulators started building in 2023 and 2024. The National Association of Insurance Commissioners adopted its model bulletin on artificial intelligence in December 2023, and by April 30, 2024, 11 states had adopted or issued related bulletins or notices, including Connecticut, Illinois, Maryland, Pennsylvania and Washington. (content.naic.org) At the same time, a startup called Claimable is using artificial intelligence from the other side of the fight. Bloomberg reported on April 22 that the company has helped reverse thousands of denied health insurance claims and has backing that includes Mark Cuban. (bloomberg.com) Becker’s, citing Bloomberg, reported that about 4,000 patients have used Claimable’s platform and roughly three in four saw their denials reversed. The service generates appeal letters for insurers and, in some cases, sends them beyond the appeals department to executives, politicians and journalists. (beckershospitalreview.com) The wider policy backdrop is getting busier, not quieter. Manatt Health reported on April 9 that 43 states had introduced more than 240 artificial intelligence bills in 2026, with payor use of artificial intelligence in prior authorization, medical-necessity decisions and claim downcoding emerging as a recurring theme. (manatt.com) That leaves insurers, regulators and vendors converging on the same question: if software helps shape a coverage decision, someone has to show how it worked, who reviewed it and how a patient can challenge it. The next wave of state rules is being written around those records, disclosures and appeals. (natlawreview.com) (content.naic.org)