Nolla opens Utah pilot for AI-generated acne prescriptions

Nolla Health has opened a Utah pilot that lets its app assess adults with mild-to-moderate acne and generate an initial prescription from a list of eight approved topical treatments. A licensed physician must review and approve every prescription in the first 100-patient phase before it reaches the patient or pharmacy. In later phases, prescriptions may be issued before physician review.

The pilot is available to Utah residents age 18 and older. It does not authorize the system to prescribe pills or hormone therapies. Nor does participation in Utah’s regulatory sandbox give an AI company a general license to practice medicine. Utah describes a sandbox agreement as a time-limited adjustment to specified rules for a named participant and task, not a state endorsement.

Patients begin with informed consent, a structured medical intake and a facial scan. Nolla’s system then analyzes acne severity and selects from the approved topical formulary. The change between phases is material: human oversight moves from advance approval of every prescription to review after prescriptions have already been issued.

The available record does not settle how often that later review takes place. Nolla’s announcement says a physician will conduct daily retrospective reviews through the first 500 patients, followed by weekly sampling. Separate reporting on the pilot describes weekly reviews for patients 101 through 500, followed by monthly reviews of at least 10% of prescriptions. The signed agreement was not located in the official record reviewed, so neither account can be treated as definitive. Utah’s general framework says clinical pilots advance on evidence rather than schedule; the available material does not establish Nolla’s specific benchmarks or whether reaching a patient-count threshold alone triggers a phase change.

Nolla says the system refers a case to a licensed physician when it cannot confidently select a treatment and lets patients message a physician at any time. These are promised operating safeguards, not independently measured results from the new pilot. The company also says clinicians agreed with its recommendations in more than 96% of real-world cases. The opened materials, however, do not provide a study protocol, sample size, subgroup results, adverse-event data or an independent audit to support that figure.

Liability and audit protections also require attribution. According to reporting on the pilot, Utah Office of Artificial Intelligence Policy director Zachary Boyd said the program requires malpractice insurance covering AI-generated prescriptions and that Nolla’s contract includes duty-of-care provisions. The reporting also says the pilot was amended to add a third-party physician appointed by the Utah Medical Licensing Board to oversee prescription reviews. Without the signed agreement or appointment record, the exact coverage, duties, authority and audit method are not independently established.

Utah’s general sandbox framework says participation does not make the state responsible for a participant’s claims or liabilities. It also says other legal obligations and patient remedies remain in force, and that the Office of Artificial Intelligence Policy may audit a participant or terminate an agreement. Those general powers are documented, but their application and results in Nolla’s pilot have not yet been shown.

The pilot is reported to run for one year, with up to two additional one-year renewals at the Utah office’s discretion. The state’s general framework independently sets the same maximum structure for demonstration periods and extensions.