Utah Lets AI Prescribe Drugs With Limited Human Oversight

The decision came from Utah's Office of Artificial Intelligence Policy, where regulators signed off on a set of new pilots that change how medication is delivered to patients. On Friday, the state approved these programs which explicitly permit artificial intelligence systems to issue and refill prescriptions without a clinician reviewing them in real time. This is not a hypothetical future scenario. It is an approved legal pathway where algorithms make the final call on what enters a patient's body.

The most visible of these initiatives involves Nolla Health, a company building a dermatology app that assesses acne from patient photos. Under this pilot, the app will test its ability to prescribe initial medications and refills from a short list of topical drugs for adults with mild to moderate acne. Initially, two physicians will review every prescription before it is sent, but in later stages, their review will happen only after the fact. This shift from pre-approval to post-approval is the critical legal and ethical leap that defines this new era of medical autonomy.

The Nolla Health Pilot

The Nolla Health pilot offers the clearest example of this autonomous shift in practice. The app uses image recognition to evaluate acne severity, a task that has traditionally required a visual examination by a board-certified dermatologist. The pilot will test whether the app can prescribe initial medications and refills from a restricted list of topical drugs to adults in Utah with mild-to-moderate acne. While the drug list is limited, adding a layer of safety, the core function remains unchanged. The machine decides the dose, the duration, and the necessity of the treatment.

The pilot is designed to reduce human involvement as it progresses. At first, two physicians will review every prescription before it reaches the patient, acting as a safety net to catch obvious errors in the algorithm's judgment. However, the end goal is to move to a model where these reviews occur only after the fact. This means that by the time a doctor sees the prescription, the patient has already received the medication. The burden of care shifts from prevention to audit, a fundamental change in the physician's role.

Chronic Care and Women's Health

The approval extends beyond skincare. Two other pilots were greenlit that address broader chronic conditions and specialized health needs. One pilot, from August AI, will have an AI system renew prescriptions for people with chronic conditions. This is a massive expansion of scope, as managing chronic diseases requires a nuanced understanding of patient history, side effects, and interactions. An algorithm handling these renewals must account for a complex web of variables that are often difficult to quantify in code.

Another approved pilot, from Expect Fitness, will offer virtual pelvic floor therapy for women in the state. This program aims to help with urinary incontinence, pelvic pain, and postpartum pelvic floor problems. While this may seem less controversial than prescribing controlled substances, it still represents a significant departure from traditional care. These conditions are deeply personal and often require a tactile, empathetic approach that is hard to replicate through a screen. The fact that regulators are comfortable allowing AI to manage these sensitive areas signals a broad acceptance of digital autonomy in healthcare.

The Regulatory Sandbox

All of these experiments are happening within Utah's Artificial Intelligence Learning Laboratory, also known as the AI sandbox. This entity was established by state law in 2024 with the explicit goal of encouraging innovation while establishing consumer protections. The sandbox provides a supervised environment to test and evaluate AI-based tools and approaches before they are scaled. Private companies that want to test their products sign an agreement with the state, which defines boundaries, safeguards, and reporting requirements. This structure allows for rapid testing that would be impossible under traditional regulatory frameworks.

The concept of the sandbox is not without its critics. Many argue that it creates a two-tiered system of healthcare, where tech-savvy users in specific states get access to cutting-edge, autonomous tools, while the rest of the country remains in a more traditional, slower-moving regulatory environment. The sandbox is designed to move fast, and it clearly is. The approval of these new pilots demonstrates that the state is willing to push the boundaries of what is considered acceptable medical practice, even if it means diverging from the norms set by the American Medical Association and other national bodies.

Pushback from Physicians

The move has not been without friction. Some of these pilots have faced ire from doctors in the state, including when a previous pilot by Doctronic was launched. Doctronic's system was designed to renew medications without clinician oversight, a move that raised immediate alarms among medical professionals. The state's medical licensing board asked that the pilot be suspended, citing concerns about patient safety and the lack of adequate human oversight. This tension between regulatory innovation and medical ethics is at the heart of the current debate.

Doctors argue that prescribing medication is not just a mechanical act of matching symptoms to drugs. It is a clinical judgment that takes into account the patient's entire context, their lifestyle, their mental state, and their ability to adhere to a treatment plan. An algorithm, no matter how sophisticated, cannot fully replicate this holistic view. The pushback from the medical community is a reminder that healthcare is a human service, and that removing the human element from critical decisions carries inherent risks that may not be fully understood until they are realized in practice.

The Road Ahead

None of the newly approved pilots have yet launched. The companies involved must still navigate the operational challenges of setting up these systems and ensuring they meet the specific requirements of the sandbox agreement. However, the fact that they have been approved is a significant milestone. It means that the regulatory hurdles have been cleared, and the path to implementation is open. The next few months will be crucial in determining whether these systems work as intended and whether they deliver the promised benefits to patients.

As these pilots move from the planning stage to active deployment, the medical community will be watching closely. The success or failure of these initiatives will have far-reaching implications for the future of healthcare in the United States. If the AI systems prove to be safe and effective, it could lead to a broader adoption of autonomous prescribing tools across the country. If they fail, it could result in a crackdown on AI in healthcare, a rollback of the sandbox model, and a reassertion of the primacy of human judgment. The outcome remains uncertain, but the experiment is now officially underway.