Utah has become the first state in the nation to allow artificial intelligence to legally participate in prescription medication renewals, launching a groundbreaking experiment that is drawing both enthusiasm and sharp criticism from the medical community. The state’s Office of Artificial Intelligence Policy announced the partnership with Doctronic, an AI health platform company, in January, with the program beginning the previous month as part of a state regulatory sandbox designed to test innovative healthcare technologies.
The pilot allows patients with chronic conditions to renew certain medications through an AI chatbot without first scheduling an appointment with their doctor. Patients verify their location in Utah and their identity using a government ID and selfie, then answer clinical questions similar to those a physician would ask. If the AI system approves the renewal, it sends the prescription directly to a Utah pharmacy for a four-dollar fee. The program covers approximately 190 commonly prescribed medications for conditions like high blood pressure, high cholesterol, diabetes and psychiatric disorders, while excluding higher-risk drugs such as painkillers, ADHD medications and injectable drugs.
State officials say the initiative addresses a genuine problem in healthcare delivery. Prescription renewals account for roughly eighty percent of all medication activity, and medication noncompliance costs the healthcare system more than one hundred billion dollars annually in avoidable expenses. When patients cannot quickly obtain refills, they often go without their medications for extended periods, increasing their risk of serious health complications such as stroke or heart attack. Proponents of the system argue that automating routine refills for stable patients would relieve clinicians of time-consuming administrative work and allow them to focus on more complex care.
However, the proposal has triggered significant opposition from physician groups and state medical regulators, who worry that removing doctors from the prescription renewal process creates unacceptable safety risks.
The Utah Medical Licensing Board sent a letter to state officials in late April expressing deep concern about the program and calling for its immediate suspension. The board emphasized that prescription refills are not routine administrative tasks but rather require clinical judgment and reassessment of whether medications remain appropriate and safe for individual patients. Each refill, the board argued, demands evaluation of potential side effects, drug interactions, and whether a patient’s condition has changed in ways that might make a medication unsuitable. The board also noted that it was not consulted before the program launched, creating what it viewed as a serious procedural failure.
The CEO of the Utah Medical Association stated flatly that the program is a bad idea, saying there are better ways to use artificial intelligence in healthcare but that AI should not be making care decisions. Medical experts raised concerns about the potential for missed drug interactions, undetected side effects, and patients remaining on outdated or suboptimal therapy for months. One chair of the Medical Licensing Board expressed fear that harm would only be discovered when deaths occurred.
The American Medical Association has also voiced caution, with its chief executive saying that while artificial intelligence has tremendous potential to improve medicine, it poses serious risks to patients and physicians when deployed without physician input. Some experts questioned whether the system adequately addresses the reality that people with mental health conditions or substance use disorders might manipulate the AI to obtain medications inappropriately.
State officials have rejected calls to halt the program, pointing to multiple safety layers built into the rollout. The pilot operates in phases, with the current phase requiring that a licensed physician review and approve every AI-generated prescription decision before it reaches a pharmacy. Doctronic was required to have the first 250 prescriptions in each medication class reviewed by human physicians before the AI could operate with greater autonomy. The company secured a malpractice insurance policy comparable to that covering physicians, and it submits detailed monthly reports to state regulators tracking approval rates, denial rates, physician reviews and any adverse events.
Company data presented to regulators showed the AI matched physician treatment decisions in 99.2 percent of cases tested in an urgent care setting. Doctronic’s cofounders have argued that for routine refills of stable patients, the AI system can be more thorough than a rushed doctor’s visit, conducting more comprehensive checks for drug interactions and contraindications than clinicians often perform in busy practices.
The system automatically escalates uncertain cases to a human physician, and patients whose prescriptions are deemed ineligible receive a code for a free video consultation with a doctor. Doctronic maintains that it has about twenty physicians on staff who review cases that raise concerns.
The program has also drawn scrutiny following a March report by security researchers who claimed to have exposed vulnerabilities in Doctronic’s artificial intelligence system. The researchers said they were able to manipulate the AI into generating unsafe clinical guidance, including incorrect medication doses. Doctronic clarified that the researchers were testing a general version of the company’s health assistant available to the public, not the prescription renewal workflow specifically designed for Utah’s pilot program.
The Utah initiative is being closely watched by other states considering similar paths. Doctronic is in discussions with states including Texas, Arizona and Missouri about launching their own AI prescription refill programs. Some observers believe the results from Utah’s year-long pilot will shape how other states and possibly the federal government approach regulating artificial intelligence in healthcare, potentially creating a national model for either encouraging such innovation or placing stricter limits on it.
The state plans to continue evaluating the pilot and retains the authority to modify or cancel the program if it fails to meet safety benchmarks. Findings will be made public to inform future state and federal artificial intelligence policy.