Utah is expanding the state's AI regulatory sandbox program with new pilots to test AI-supported prescription renewals, dermatology and physical therapy, while adding physician oversight and independent third-party evaluators to strengthen safety and accountability.
Utah is bringing the Utah Department of Health and Human Services, Intermountain Health, University of Utah Health, medical licensing officials and outside experts into the state’s AI policy process. State officials amd health systems leaders say the initiative is designed to test AI in real-world clinical settings, generate evidence on its benefits and risks and address key questions around liability, privacy and regulation.
At a press conference at the State Capitol on Monday, state officials announced a new clinical AI advisory group and network of independent third-party evaluators to audit healthcare AI sandbox pilots. Initial evaluators include Coalition for Health AI (CHAI), Clarion AI Partners, Glacis Technologies, mpathic AI, Vega Health and Stanford Clinical Excellence Research Center.
These technical and clinical partners will stress-test system guardrails, inspect decision logs and evaluate patient outcomes alongside state regulators. Bringing clinical evaluators into the process ensures that claims made by AI developers are independently verified, state officials said.
Utah’s closely-watched AI regulatory sandbox program is a first-of-its-kind state initiative administered by the Utah Department of Commerce's Office of Artificial Intelligence Policy (OAIP). The sandbox creates a structured, supervised environment where businesses can deploy and test innovative AI products under a tailored regulatory framework. In January, the state inked a partnership with startup Doctronic to test out an AI-powered system for rescription renewals as the first healthcare AI pilot in the sandbox program.
State officials stressed that the use of AI in healthcare could improve access to care, lower the cost of medical care, support clinicians' work and expand the capacity of the healthcare workforce.
The state's healthcare AI regulatory sandbox positions Utah as a national test bed for AI regulation that balances innovation with patient safety. state officials said. The aim of the regulatory sandbox is to "let the market learn and the state learn together," said Zach Boyd, director of the Office of AI Policy, during the event at the State Capitol on Monday.
"We're taking a market-based approach because we want to support a validation landscape that will grow as the technology grows and the capability expands," Boyd said.
On Monday, the Office of Artificial Intelligence Policy announced five new healthcare AI sandbox pilots with August AI, Nolla Health and Expect Fitness. August AI will pilot an AI-assisted prescription renewal program for individuals with chronic illnesses while Nolla Health uses AI to evaluate patients for acne treatment. Expect Fitness will pilot AI-enabled pelvic floor physical therapy services.
August AI’s pilot covers routine, non-controlled medications for chronic conditions, including the most commonly prescribed first-line treatments for diabetes and cardiovascular disease. The company will charge patients $4 per prescription renewal. Utah-licensed physician will review every renewal in the pilot before authorizing the prescription, the company said. Its AI health companion is used by more than nine million people across 160 countries.
The pilot is stage-gated by drug class, and August must meet performance requirements at each stage before it can move to the next. The pilot also includes the safeguards required of every participant in Utah's sandbox as patients are told when AI is involved in their care and August carries specialty medical malpractice insurance. Prescriptions are issued under the licenses of August's Utah-licensed physicians.
As part of Nolla Health's pilot, consumers in Utah can use the Nolla Derm app to get initial acne treatment prescriptions for $4.99 a month. Each patient verifies their identity, goes through informed consent, a short intake questionnaire and a quick five-angle face scan using the AI-powered app. The AI generates a personalized treatment plan and, if appropriate, an initial prescription. Nolla says clinicians agree with the AI’s treatment recommendations in over 96% of real-world cases. Physicians oversee every stage of the rollout, according to the company.
The Office of AI Policy also approved "umbrella agreements" with the University of Utah Health and Intermountain Health, which set the terms for approving future AI pilots.
As health systems face staffing constraints and ongoing demand for medical services, AI can be an important tool to "bridge from an old model to a new model," said Dan Liljenquist, chief strategy officer at Intermountain Health, during the press conference.
"We're determined to lean in safely and in appropriate ways to see how AI can wrap around our providers to remove friction, to improve coordination, to actually help us provide better healthcare," Liljenquist said. "That has to be done in a safe, responsible way, and this AI sandbox partnership with so many others who are leaning in to help test and to make sure that AI is safe for use, that's important, and it can't be done in a vacuum."
He stressed that the goal is to use AI to augment, not replace, clinical judgment with a focus on keeping the clinician-patient relationship at the center of care.
Intermountain currently has 300 AI intiatives underway, mostly focused on administrative functions, he noted. "What's unique about this sandbox is this will give us the ability to lean in on those areas, those edges that require licensure from the state, and we're very excited to test some of those areas to see if we might remove some of the burden that is upon our caregivers today to again meet the demand with limited resources," Liljenquist said.
Utah is leading the nation in creating the environment with smart regulation that unlocks innovation, said Nirav Shah, M.D., senior scholar at Stanford University's Clinical Excellence Research Center, during the event.
"We're figuring it out together with industry, with regulators, and setting the stage for what I think is going to be the national model for how AI is regulated, and part of that is to understand evaluation, to make sure that these tools work as they claim to," Shah said.
Earlier this year, the Utah Medical Licensing Board raised concerns about the Doctronic pilot, saying that the board was not consulted about the pilot program and that autonomous or semi-autonomous AI refills bypass clinical safeguards. The board called for more professional medical oversight of the healthcare AI pilots.
The board now has a working group that will work with the AI policy office to vet the healthcare AI applications.
The physician-led working group will evaluate healthcare AI pilots, set evidentiary standards and provide ongoing oversight, accoding to William Hamiton, M.D., an anesthesiologist, medical board member and a member of the new working group.
The working group is not a blanket board of AI pilots, Hamilton noted. Authorized pilots will be controlled by multi-phase safety checks, physician oversight and continuous data auditing.
The collaboration between the AI policy office and the board's working group has four key objectives—physical acquisition-led oversight, rigorous safety protocols, patient-centric protections and gathering real-world data to guide future policy and regulations.
Using the evolution from slide rules to calculators as an analogy, Hamilton argued that healthcare is at a similar technological turning point with AI. "Utah can lead the world in safety and in implementing this amazing technology," he said.
The use of third-party evaluators creates another safeguard as Utah continues to test healthcare AI. These evaluators will help ensure that the tools in the sandbox meet the Office of AI Policy's evidentiary standards, noted Mark Sendak, M.D., CEO of Vega Health,
"Third-party evaluators will independently assess AI products, while all decisions about what to include in the sandbox will remain with the state," he said during the press conference.
AI companies, health systems and regulators all play a role in evaluating AI tools, but independent third-party assessments are essential for validating claims of public benefit. he noted. "Evaluations by the third-party evaluators will match the specific use cases being tested. Evaluators will ask questions like, 'What benefit is expected? What are the failure modes? What evidence would change our decision to approve or not approve a product?' That rigor matters most for clinical AI," said Sendak, who spent over a decade at the Duke Institute for Health Innovation building clinical AI.
Vega Health is well positioned for the role as a third-party evaluator because the company has experience conducting independent model evaluations and generating evidence in support of health systems and their priorities, Sendak noted.
Utah's sandbox is one of the first places where that evidence is going to be generated, Sendak said. "These independent evaluations will surface valuable learnings for clinicians who want to better understand how to apply AI in their practice, companies who are trying to build and improve their products, regulators who are making the decisions about what to include for sandbox participation, and ultimately for the people of Utah, we can feel more confident that AI in the sandbox has been extensively evaluated," he said.
Lawmakers said the sandbox provides a real-world testing ground for medical AI to help address key questions around liability, accountability, patient privacy and transparency. The results of the AI pilots will help shape future legislation, lawmakers said.