States Tighten Rules on AI Health Coverage Decisions

PYMNTS, summarizing a Sheppard Mullin Healthcare Law Blog review, reported that several states are expanding guardrails on AI use in health insurance decisions. Sheppard Mullin says Georgia's SB 444, effective January 1, 2027, allows AI to participate in decision processes but bars an AI system from issuing an adverse determination until a qualified human reviewer conducts utilization review involving a clinical peer. The same review says Iowa's HF 2635, effective July 1, 2026 for the AI provision, permits AI for initial prior authorization review but bars AI as the sole basis for denial, delay, or downgrade based on medical necessity.

The AMA's state legislative review lists enacted measures in states including Arizona, Illinois, Maryland, Nebraska, and Texas that constrain healthcare AI use through requirements such as disclosure, human judgment, non-discrimination, accuracy review, and limits on replacing clinical decision-making. A Colorado Legislative Council memo citing NCSL documents the wider surge in health-related AI bills, which shows why national payer workflows are likely to face a patchwork of state requirements.

The engineering requirement is stronger governance plumbing. Systems used in prior authorization or utilization review should capture model version, input evidence, output rationale, human reviewer identity or role, reviewer notes, final decision, appeal context, and fairness-monitoring signals. If the system cannot show how a human clinical reviewer independently assessed the recommendation, the model may be hard to defend under these laws.

Watch whether states move from process controls to stricter limits on model classes, automated-denial workflows, or training data. Also watch for federal preemption debates and insurer adoption of standardized decision packets, because those could determine whether engineering teams build one national audit pattern or many jurisdiction-specific workflows.