0 human doctors on board: AI chatbot Doctronic begins prescribing in Utah pilot

For AI developers and researchers, Doctronic represents a historic milestone: a large language model-based chatbot has been entrusted with real-time, autonomous prescription refills—a function that until now required a medical license. With the program poised to drop human doctor review entirely, the question of how to validate, explain, and hold accountable an AI’s clinical decisions moves from theory to immediate regulatory reality.

A prescription refill pilot program in Utah has thrust artificial intelligence into a contentious medical debate: whether an AI chatbot can be entrusted with the direct authority to prescribe medication. The program, which quietly launched earlier this year, allows residents to bypass a physician’s visit and obtain prescription refills through an AI chatbot called Doctronic. It marks the first known instance in the United States where a non-human entity has effectively been granted a form of prescribing license—albeit under the umbrella of a state regulatory sandbox that temporarily waives longstanding medical practice laws. This development shatters a century-old legal precedent that reserves the act of prescribing solely for licensed medical professionals, and it raises profound questions about patient safety, regulatory oversight, and the future role of AI in clinical decision-making.

“

This would attract venture capital and tech giants eager to capture a slice of the $500 billion U.S.

At the operational core, Doctronic currently operates with guardrails: during its initial phase, every AI-generated refill order is reviewed by a human doctor. However, the company behind the pilot expects to transition soon to fully automated refills, eliminating human oversight entirely. The oversight board consists of five AI specialists, none of whom are medical doctors, underscoring a fundamental shift in how medical authority is defined and delegated. Proponents argue that AI can streamline routine care, reduce costs, and improve access, especially in underserved areas where physician shortages are acute. But critics—including doctors, legal experts, and public health advocates—contend that no AI system has undergone the rigorous, standardized training, testing, and ethical vetting required of human physicians.

The regulatory mechanism enabling this experiment is Utah’s ‘regulatory sandbox,’ a program that allows AI companies to test innovative technologies by temporarily waiving certain laws. While such sandboxes have been used in fintech and other sectors, their application to healthcare is unprecedented and fraught with unique risks. The most immediate concern is patient safety: prescription errors, drug interactions, and contraindications can have deadly consequences. The lack of physician involvement in the oversight structure exacerbates fears that clinical nuance will be lost. Legal liability is another murky area—if an AI-generated prescription harms a patient, it is unclear who is accountable: the software developer, the sandbox administrators, or the state.

Dr. Eric Bressman of the University of Pennsylvania captured the sentiment of many medical professionals when he said, “We have crossed a threshold in terms of giving something that is not human a medical license, whether or not we want to call it that.” His remark highlights the symbolic and practical departure from established norms. The American Medical Association and state medical boards have so far remained publicly cautious, but behind the scenes they are mobilizing to challenge the sandbox’s legality under federal and state laws that govern the practice of medicine.

The pilot’s outcome could have far-reaching implications for the healthcare and AI industries. If Doctronic proves safe and effective, other states may adopt similar sandbox models, accelerating the deployment of AI in routine care and creating a new market for autonomous prescribing platforms. This would attract venture capital and tech giants eager to capture a slice of the $500 billion U.S. prescription drug market. Conversely, a high-profile failure could trigger a regulatory backlash that stalls AI integration in clinical settings for years. Already, patient advocacy groups are calling for transparency, mandatory adverse event reporting, and independent auditing of the AI’s decision-making logic.