Utah Lets AI Renew Prescriptions. Doctors Want the Program Suspended.

What the Program Actually Does

Utah's AI prescription pilot is not replacing doctors. Patients still need a physician to write an initial prescription before the system touches anything. What the program does is automate refills — the repetitive, guideline-based renewals that stack up in clinic queues and slow down care for everyone.

The platform is powered by Doctronic, an AI startup, and administered by Utah's Commerce Department. A separate agreement covers Legion Health, an online platform using AI to process low-level mental health medication refills. Both operate under the state's Office of Artificial Intelligence Policy, which was created specifically to run exactly these kinds of experiments.

According to the Office, the goal is to "lighten clinician workload, improve refill continuity, and expand patient access — all while maintaining the same safety and quality standards patients expect."

Whether it holds up in practice is a legitimate open question.

Who's Objecting and Why

The Utah Medical Licensing Board sent a letter to the Office of Artificial Intelligence Policy in April asking the program to be suspended immediately. The Office declined, asserting it has the statutory authority to continue.

Alan Smith, a family physician who chairs the Utah Medical Licensing Board, told The Wall Street Journal that the liability question is unresolved. "People can have life and death reactions to medications," Smith said. "Who is actually liable for problems that may occur because of a refill of a medication?" Smith noted he was speaking for himself, not the board — though the board's April letter makes clear the concern is shared.

John Whyte, chief executive of the American Medical Association, raised concerns about a slippery slope. "It's really a concern about a slippery slope. Do we start here with prescriptions, then all of a sudden, is it going to lead to diagnostic tests?" Whyte told the Journal. "Medical decisions are complicated, they're nuanced, and it's the fundamental component of the physician-patient relationship."

These are fair concerns, not reflexive technophobia. Liability in AI-assisted medicine is genuinely unsettled law. If an AI renews a prescription and the patient has a serious reaction, the question of who bears responsibility — the startup, the state, the original prescribing physician — has no clean answer in current tort doctrine.

The Strongest Case for the Critics

The medical establishment's concern deserves a straight hearing before it gets contextualized. Prescription refills sound routine until they aren't. Drug interactions change when other medications are added. Patients develop new conditions. Dosages that were appropriate six months ago may not be appropriate today. A physician doing a refill review is supposed to catch those changes. An AI system trained on guidelines may not.

For mental health medications in particular — the ones processed through Legion Health — the stakes are higher. Critics in the medical field contend that AI cannot replicate the nuanced clinical judgment required for mental health treatment. Psychiatric drugs have narrow therapeutic windows, significant withdrawal risks, and complex interactions with a patient's life circumstances. Those critics are not wrong as a general matter.

The question is whether that concern applies specifically to routine refills of stable, well-understood conditions under established guidelines or whether it's being used to block a narrower use case that might actually be safe.

What the Evidence Shows So Far

Early results from the Doctronic pilot are notable. The chatbot has refilled prescriptions 72 percent of the time without escalating to a physician for reasons like missing lab work, according to STAT. Physicians agreed with the program in 91 percent of cases where the chatbot granted a renewal. Doctronic also cites a clinical study finding that its top diagnosis matched a clinician's in 81 percent of cases, and that treatment plans aligned in 99.2 percent of cases.

Still, significant uncertainty remains. As STAT points out, it is not clear an AI bot can legally renew prescriptions without FDA authorization. The FDA has not provided clear regulations on AI, and if it determines chatbots like Doctronic qualify as medical devices, it could block platforms like this from the market entirely.

The Office of Artificial Intelligence Policy tells Reason it is also "exploring the role of third-party auditors to help develop a standardized auditing process that includes transparent reporting and increased trust." Work remains to be done on these safeguards.

The Real Unresolved Question

Both sides are talking past each other on scope. The Office is defending a system designed for routine, guideline-based renewals of stable conditions. The AMA and the Medical Licensing Board are raising concerns about where AI might go — diagnostic authority, complex clinical decisions, mental health treatment.

Those are different arguments about different things. The pilot as described is narrower than its critics suggest and less proven than its defenders claim.

What would settle this: transparent, third-party-audited outcome data from the existing pilots, with clear liability language written into state statute. The Office says it is exploring third-party auditing. Whether Utah's legislature moves to codify liability rules — or whether the Medical Licensing Board escalates its push for suspension through formal regulatory channels — is the concrete next step that will determine whether this experiment continues or collapses under its own legal uncertainty.