Utah Expands AI Healthcare Experiment Beyond Prescription Renewals

October 6, 2026 by matray

Utah regulators are expanding a groundbreaking experiment that permits artificial intelligence to perform clinical functions traditionally reserved for licensed healthcare professionals, moving beyond prescription renewals into new prescriptions and treatment planning while building a broader regulatory framework for the technology.

The latest approvals show that Utah’s first AI prescription-renewal experiment with Doctronic was not a one-off. The state is now applying variations of the model to additional companies and clinical settings, including chronic-disease prescription renewals, acne treatment and pelvic-floor physical therapy.

Among the newest participants is August AI, which has been approved by the Utah Department of Commerce’s Office of Artificial Intelligence Policy to test AI-authorized prescription renewals for adults with chronic conditions.

The pilot initially requires intensive human review but could eventually allow August’s AI to authorize qualifying prescription refills while licensed providers review only a sample of its decisions.

The August program builds on Utah’s earlier Doctronic prescription-renewal pilot (see MLM, February 2026), the first state-authorized program to test AI-assisted prescription renewals through Utah’s regulatory sandbox.

Doctronic uses AI to process renewals for previously prescribed medications. Its program attracted national attention because it established a pathway under which AI could eventually authorize qualifying prescription renewals without prospective physician approval of every decision.

The Doctronic program drew objections from the Utah Medical Licensing Board, which warned against treating prescription renewals as simple administrative transactions, arguing that physicians use refill requests to reassess patients, monitor medications and identify changes in their conditions.

The newly approved August pilot applies a similar concept through a defined, stage-gated review process. Under the program, the AI can authorize routine 30-, 60- and 90-day refills of maintenance medications that adult Utah patients are already taking under established treatment plans. The program is limited to a predetermined list of noncontrolled medications and cannot be used to initiate a prescription, change a dosage or substitute one medication for another.

A Utah-licensed provider must prospectively review the first 250 refills before they reach patients. Providers will retrospectively review each of the next 1,000. Advancement requires written approval from the Office of Artificial Intelligence Policy.

If August meets the state’s performance requirements and is permitted to advance, providers would eventually spot-check at least 5% of refills in each medication class rather than individually reviewing every renewal. The demonstration period has not yet begun.

Utah regulators granted August limited regulatory relief to conduct that experiment. The state’s Division of Professional Licensing agreed not to pursue certain enforcement actions for unlicensed or unprofessional conduct or under state telehealth requirements when the AI authorizes qualifying refills in accordance with the approved protocol.

The pilot also includes safeguards designed to limit the AI’s autonomy. According to the Office of Artificial Intelligence Policy, August must verify a patient’s identity and pharmacy record before issuing a refill. Initially, patients may receive no more than three consecutive AI-authorized refills of a medication before seeing a human provider. Signs of suicidal thinking, new side effects, potentially dangerous drug interactions or missing laboratory monitoring automatically stop the AI refill process and send the case to a licensed provider.

The program also carries provisions with direct medical liability implications. Companies participating in Utah’s healthcare AI pilots must maintain medical professional liability insurance coverage, disclose the use of AI to patients and comply with adverse-event reporting requirements. Utah’s agreements do not eliminate companies’ liability for patient harm, and regulatory relief is narrowly limited to the activities authorized by each pilot.

The latest approvals also extend Utah’s regulatory model beyond the prescription renewals first tested with Doctronic.

A potentially more consequential pilot involving Nolla Health will test AI prescribing of topical acne medications. Unlike August and Doctronic, which deal with medications previously prescribed as part of an established treatment plan, Nolla’s system can assess qualifying adult patients and issue an initial prescription for certain topical acne treatments.

The Nolla program will also proceed in stages. Two Utah-licensed physicians will review every prescription before it is sent during the first stage. Later stages can move to retrospective review and eventually sampling, but each advancement requires written approval from the Office of Artificial Intelligence Policy. The Nolla demonstration period also has not yet begun.

Another approved program, Expect Fitness, moves the model beyond prescribing altogether by testing AI-generated pelvic-floor physical therapy plans for adult women. A licensed physical therapist initially must approve every plan, but the pilot could eventually allow lower-risk plans to reach patients subject to provider spot checks.

Taken together, the programs demonstrate a common model emerging in Utah: AI systems begin with comprehensive human oversight and can receive progressively greater clinical autonomy after generating performance data and receiving additional state approval.

As the number and scope of the pilots grow, Utah is adding another layer of oversight through an independent third-party evaluation network designed to verify claims made by participating companies.

Six outside organizations have entered agreements with the Office of Artificial Intelligence Policy to provide various forms of clinical or technical evaluation, evaluator-network services or independent verification that required safety checks occurred.

Evaluators can test systems for clinical errors and unsafe behavior before a pilot begins, review privacy safeguards and audit samples of actual AI activity after deployment. They can also compare companies’ performance reports against the underlying activity.

As Utah’s experiments progress from AI-assisted decisions toward potentially AI-authorized clinical care, their results could help answer questions extending well beyond the state: when AI can safely assume a clinical function, how much human oversight is necessary and who bears liability when an AI-assisted or AI-authorized healthcare decision causes patient harm.