On Monday, October 5, 2026, New York-based health startup Nolla Health announced that it has become the first organization in the United States authorized by a state regulator for an artificial intelligence system to issue initial medical prescriptions, not just refills. The program starts in Utah through the Nolla Derm app and is limited to acne treatments.
The authorization follows months of work with Utah's Office of Artificial Intelligence Policy, which runs a regulatory sandbox for new technologies. According to the company, the pilot went through several rounds of external review by practicing physicians, dermatologists and members of the state government, and Nolla Health met with the Utah Medical Board before launch to incorporate its feedback.
The process is simple for the user. Utah residents aged 18 and over download the app and enroll in the program for a discounted fee of $4.99 per month; medications are paid for separately. After verifying their identity, completing informed consent, answering a 10- to 15-minute questionnaire and taking a five-angle face scan, they receive a treatment plan and, if appropriate, a prescription.
The system is tightly bounded. It only serves adults with mild-to-moderate acne who can benefit from a specific list of topical treatments; according to Bloomberg, the AI can currently prescribe eight different skin treatments. Severe acne and other conditions are screened out, and if the system cannot “confidently select” a treatment, it refers the patient to a licensed physician.
Physician oversight is designed to step down in stages. Under the signed Utah agreement as described by the company and by MobiHealthNews, in the first stage two Utah-licensed physicians review and approve every AI-generated prescription before it reaches the patient, for at least the first 100 patients. In the second, up to 500 patients, the AI issues prescriptions directly and a physician reviews each case after the fact.
In the third stage, physicians review a sample of at least 10% of prescriptions each month, plus every case involving side effects or escalation. Moving from one stage to the next requires meeting safety targets, including 95% agreement with physicians and zero serious adverse events, as well as written approval from the state. The company's press release describes a slightly different review cadence, which is worth watching as the state publishes more details.
Nolla Health justifies the project with access data. According to the company, about 100,000 people in Utah have active acne at any given time, but only 10% ever see a dermatologist. The average wait in the state is 61 days, 11 counties have no dermatologist at all and a visit can cost up to $300. The company says clinicians agree with its AI's recommendations in more than 96% of real-world cases.
“AI will matter most in healthcare when it helps people get care they otherwise would not receive,” said Luis Wenus, co-founder and CEO of Nolla Health. Zach Boyd, director of Utah's Office of Artificial Intelligence Policy, called the pilot a significant step toward improving healthcare access while maintaining rigorous oversight.
The company's chief medical advisor, Zaid Fadul, defended the system's limits: “The AI can't improvise. It can only choose from a short list of physician-approved topical treatments.” He added that the real test for AI in medicine is whether someone other than the company can shut it off: in Utah, he said, both the state and physicians can.
The company stresses that the pilot is meant to complement physicians, not replace them, by handling routine, low-acuity cases and freeing dermatologists for higher-risk patients. Patients can message a licensed physician through the app at any time. According to Nolla, the app has been downloaded more than 175,000 times across 44 states.
The launch is not without controversy. Bloomberg Law noted that the program comes months after Utah's efforts to expand AI's role in healthcare drew scrutiny and anger among doctors. The Verge pointed out that the state had already allowed AI systems to renew certain medications earlier this year and that several companies are experimenting with AI-generated prescriptions within that framework.
The case will serve as a national reference point. If the pilot meets its safety targets, other states could study similar models for common, low-risk conditions. If problems emerge, it will strengthen the argument of those who want prescribing to remain an exclusively medical act. Either way, data from the early stages will be decisive.
In short, Utah is opening an unprecedented door in the United States: an AI that issues initial prescriptions under strict rules and declining oversight. The scope is limited, the treatment is low-risk and the regulator keeps control, but the precedent matters. What happens with the first 100 and 500 patients will determine whether this model expands or stalls.
