The Utah Medical Board Just Called for Suspension of the State’s AI Doctor Experiment. The Reasons Are Unsettling.

Like most American healthcare experiments, it began quietly with a press release in early January, a few cautious paragraphs about innovation, and a statement that ought to have sparked more attention than it did: the first AI to legally prescribe routine refills.

The agreement between Utah and Doctronic, a venture-backed company with over $65 million in funding, was presented as a cautious, closely monitored move into the future. By April, the state’s medical board was calling for its closure.

The letter, dated April 20 and signed by eleven clinicians, was blunt. Depending on your point of view, the Utah Medical Licensing Board’s statement that it only found out about the pilot after it went live could be interpreted as either a purposeful workaround or an oversight. Reading between the lines, the tone of the board points to the latter. “We must not allow AI or other financial motivations to override this obligation,” the physicians wrote, “yet that is precisely what occurred here.” That is not how bureaucratic disputes are expressed. People who feel excluded from a decision they feel was theirs use that language.

Observing the entire episode gives the impression that two radically different medical theories are colliding in real time. A startup pitching efficiency, a state office of AI policy, and a regulatory sandbox intended to allow innovative technologies to avoid the typical obstacles are all on one side. Conversely, a board of doctors consistently reiterates the outdated notion that a prescription refill is not a transaction.

The choice is clinical. Doses fluctuate. The side effects start to appear. Patients undergo changes without disclosing them. The woman renewing her blood pressure medication has started two new medications from a cardiologist in the last six months, but a bot retrieving data from a database is unaware of this.

The framing used by Doctronic has been clever—perhaps too clever. Co-founder Matt Pavelle has spent months referring to his product on LinkedIn as “the only licensed AI doctor,” despite the fact that the company’s bot isn’t authorized to practice medicine, which would draw federal scrutiny. Technically, the prescriptions are written in the name of a legitimate Utah-licensed physician. This brings up the question Michelle McOmber of the Utah Medical Association frequently poses: whose name is on the lawsuit if something goes wrong? “That person would be liable,” she replied, “if anything goes wrong.” It’s difficult to ignore how infrequently that question appears in the upbeat press materials.

The bot hasn’t really renewed anything on its own yet. A doctor must review the first 250 prescriptions in each drug category during the pilot’s first phase, and the threshold hasn’t been reached. Although the Utah AI office has not yet released the promised statistics, it claims that no significant safety incidents have been reported. Disagreements between the AI and its supervising clinicians have been documented in the office’s euphemistic taxonomy as “no risk,” “minor risk,” or “minor-to-moderate risk.” It’s the kind of phrase that sticks around, whatever that final category actually means.

Doctronic intends to grow into Texas, Arizona, and Wyoming—states with comparable regulatory flexibility. It remains to be seen if Utah’s pilot will survive the medical board’s intervention. The more complex question, which no one at the launch event seemed eager to address, is whether convenience should be prioritized when someone’s heart medication is at stake. Technology advances quickly. For reasons that the board has now documented in writing, medicine typically does not.