Utah is building a verification mechanism for its healthcare AI regulatory sandbox, a structure that allows AI health tools to operate under state oversight while real-world safety and efficacy data is gathered — an approach that stands as a distinct regulatory model compared to guideline-based frameworks like the one South Korea recently issued for mental health AI specifically.
How Regulatory Sandboxes Work
A regulatory sandbox typically allows companies to deploy new technology under closer-than-normal state supervision and relaxed certain regulatory requirements, in exchange for enhanced reporting and data-sharing obligations that let regulators learn how the technology performs in real-world conditions before deciding whether and how to regulate it more permanently. Utah’s approach to healthcare AI specifically reflects growing interest among state governments in finding middle paths between prohibiting new AI health tools outright and allowing unrestricted deployment without any data-gathering phase.
The Verification Mechanism’s Purpose
The verification system Utah is building would presumably assess whether AI health tools operating within the sandbox are performing as claimed and meeting whatever safety benchmarks the state has established for sandbox participation — a function that requires technical expertise most state health regulators have historically lacked, given how recently sophisticated AI verification needs have emerged as a regulatory requirement.
How This Compares to Federal Approaches
Utah’s sandbox model operates somewhat independently of federal AI health regulation, including the FDA’s device clearance pathways and ongoing discussion-paper process for generative AI-enabled medical devices, open for comment until October 19. States building their own frameworks alongside federal regulatory development has created a genuinely multi-layered regulatory landscape for health AI companies to navigate, one that varies significantly depending on which states a company wants to operate in and what federal pathway, if any, its product requires.
Why States Are Moving Independently
State-level health AI regulatory experimentation has accelerated partly because federal AI policy has moved unevenly and inconsistently across different administrations and congressional sessions, leaving states like Utah to develop their own frameworks rather than waiting for federal clarity that may not arrive on a predictable timeline. This mirrors a broader pattern across U.S. AI policy generally, where states including Colorado, Connecticut and Maryland have each passed their own distinct AI regulations covering different sectors and use cases.
Industry Reaction
Health AI companies have generally expressed cautious interest in sandbox models like Utah’s, since they offer a path to real-world deployment and data generation without the full regulatory burden of immediate comprehensive compliance — though companies operating across multiple states still face the challenge of navigating different sandbox or regulatory frameworks in each jurisdiction where they want to operate.
What’s Next
Utah has not published a completion timeline for its verification mechanism or detailed which specific AI health tools are currently operating within the sandbox. Health policy researchers will likely watch Utah’s experience as a potential model — or cautionary tale — for other states considering similar regulatory sandbox approaches to health AI oversight.
Photo: Brett Sayles / PEXELS via Pexels
