Utah Is Letting an AI Chatbot Prescribe Psychiatric Drugs Without a Doctor in the Loop

Utah has authorized an AI system to prescribe psychiatric medications — including controlled substances — without real-time physician oversight, making it only the second US state to delegate this level of clinical authority to an automated system. The program is now operational, and it is dividing the medical community sharply.

The AI system, deployed through a telehealth platform approved by the Utah Division of Professional Licensing, handles the full prescribing workflow: patient intake, symptom assessment, medication selection, dosage, and refill management. A licensed physician reviews cases asynchronously, but is not required to be present during the consultation or to approve prescriptions before they are issued.

The Case for It

Utah's argument is straightforward: the state has a severe shortage of psychiatrists, particularly in rural counties. Wait times for a first psychiatric appointment can exceed four months in some areas. The AI system can see patients immediately, at any hour, for a fraction of the cost. State officials point to early data suggesting the platform has significantly improved medication adherence among patients who previously went untreated due to access barriers.

"We're not replacing psychiatrists," one state health official told the Salt Lake Tribune. "We're providing access to a baseline level of care that didn't exist before for tens of thousands of Utahns."

The Case Against It

Physicians are less sanguine. The primary concern is not the AI's diagnostic capability in straightforward cases — it is the system's behavior in atypical presentations, edge cases, and the complex drug-interaction scenarios that characterize psychiatric practice.

"Psychiatry is not pattern matching," said one physician who requested anonymity due to institutional restrictions. "A patient presenting with what looks like depression might actually be in a prodromal phase of bipolar disorder, where an SSRI without a mood stabilizer can trigger a manic episode. Catching that requires context, observation, and a kind of intuition that doesn't reduce to rules."

The opacity of the underlying system is a secondary concern. The platform's prescribing logic is proprietary, meaning neither patients nor reviewing physicians have access to the reasoning chain behind a given prescription decision. When things go wrong, accountability is murky.

The Regulatory Landscape

The FDA currently regulates AI diagnostic tools as medical devices but has not established a clear framework for AI-driven prescription systems. Utah's program exists in a regulatory gray zone that has so far not attracted federal intervention — though physicians' groups have petitioned the FDA to classify autonomous prescribing AI under the Class III high-risk device category, which would require pre-market approval and ongoing safety monitoring.

The outcome of Utah's experiment is likely to influence how other states approach the question. With AI systems demonstrably capable of handling routine psychiatric cases at scale, the pressure to replicate the program in shortage states will be significant — especially as healthcare systems face mounting cost and access pressures. The question is whether the regulatory infrastructure will keep pace.