Utah has partnered with six independent evaluators to check healthcare AI pilots’ safety and performance before and during state-supervised deployment.
Utah is adding independent evaluation to its healthcare AI regulatory sandbox, requiring participating tools to undergo outside review rather than relying only on developers’ own performance claims. The state’s Office of Artificial Intelligence Policy has agreements with six organizations: the Coalition for Health AI, Stanford Clinical Excellence Research Center, Clarion AI Partners, Glacis Technologies, Vega Health and mpathic AI.
The evaluators will review technical and clinical claims, audit AI decision-reporting accuracy and monitor pilots as they operate. Utah officials retain authority over whether a project starts, continues, changes or advances. Participating companies contract with and pay evaluators directly; the state says evaluators must report patient-safety concerns and permit audits of their work. The framework calls for evidence from real-world use, not just pre-deployment testing.
The oversight approach follows scrutiny of an earlier sandbox pilot for automated prescription refills. In April, Utah’s medical licensing board asked the state to suspend that pilot over concerns that patients could continue refilling medications without necessary clinical reassessment. The state says it has since strengthened reporting requirements and changed its review process.
Alongside the evaluator network, Utah announced new healthcare pilot agreements with University of Utah Health, Intermountain Health, Nolla Health, Expect Fitness and August AI. Proposed uses include acne-treatment pathways, postpartum and pelvic-health support, and refills for routine, non-controlled chronic medications. The refill pilot is subject to human physician review before prescriptions are authorized. The agreements also include patient disclosures, incident-response requirements and state oversight.
For healthcare providers and AI vendors, the model makes independent evidence and ongoing monitoring part of the path to testing clinical tools in a state sandbox. It does not amount to a general certification that the products are safe or effective, nor does it establish a nationwide standard. Utah says results from the pilots will help inform future state rules. The program’s significance will depend on what its evaluations find and how regulators respond when performance or safety falls short.