Utah allows Legion Health chatbot to prescribe psychiatric medications

Utah licensed a chatbot from Legion Health startup to renew prescriptions for psychiatric medications without mandatory doctor participation. This is only the second case in US history when clinical authority has been officially transferred to an artificial intelligence system. The year-long pilot project, announced last week, operates on a simple scheme: patients in Utah subscribe for $19 per month, communicate with Legion Health's chatbot, and in certain cases receive updated prescriptions for psychiatric medications without consulting a live psychiatrist or therapist.

The San Francisco-based company positions the product as a way to get "fast and simple" prescription renewals. Utah authorities cite two main goals for the experiment: reduce the cost of psychiatric care and partially address the acute shortage of specialists. The US faces a critical shortage of psychiatrists: according to the Association of American Medical Colleges, the shortage of mental health physicians could reach 30,000 by 2036.

In such a situation, delegating routine tasks to an algorithm seems like a reasonable compromise to authorities. The medical community responds with skepticism. Doctors point to several problems.

First, the chatbot is a black box: it's unclear on what data and logic it bases clinical decisions and how it tracks side effects or drug interactions. Second, psychiatric medications are not routine drugs: antidepressants, antipsychotics, mood stabilizers require regular direct assessment of the patient's condition. Third, according to critics, the $19/month subscription model is more likely to serve those who already have healthcare access than to help the poor, uninsured, and residents of remote areas.

It's important to understand the scale: this is about renewing already existing prescriptions in certain cases, not about initial diagnosis. Nevertheless, this is a clear shift in regulatory logic—American states are increasingly willing to test AI in clinical practice without waiting for federal standards. A key question remains unanswered: who is responsible if the chatbot renews a prescription for a patient who shouldn't have it?

What is the escalation procedure if the system "sees" concerning symptoms? Legion Health does not disclose details of its architecture and safety mechanisms. If the year-long pilot completes without serious incidents, this will be a strong argument for regulators across the country.

Psychiatry is particularly sensitive to errors: the cost of a wrong decision here is measured not in discomfort but in crises and hospitalizations. To delegate this area to AI without strict transparency standards means building trust on credit that the system hasn't yet earned.