Utah Opens the Door to AI Prescribing for Acne Care
Utah has become the first state in the US to permit artificial intelligence to issue new prescriptions, beginning with acne drugs for adults with mild to moderate disease. The approval marks a sharp regulatory shift because the system is being built to move beyond physician review and, eventually, toward independent prescribing by software. For now, the company behind the program says every prescription still passes through two Utah-licensed doctors before it is issued.
The platform at the center of the decision is Nolla Health’s Nolla Derm app. Adults in Utah with mild to moderate acne can submit facial scans and medical information for algorithmic review, after which the system recommends treatment. In its first phase, the workflow still includes physician oversight, but the structure is intentionally designed to evolve. The company says the agreement includes performance requirements and safeguards that must be satisfied before progression to later phases.
The clinical and regulatory meaning of the move is unusually broad for a condition that many people consider routine. Acne is common, often chronic, and frequently managed through repeat medication decisions rather than one-time intervention. In that setting, the promise of automated triage is obvious: faster access, lower friction, and fewer bottlenecks for patients who might otherwise wait for an appointment just to renew or initiate therapy. Yet acne prescribing is not trivial. Treatment choice depends on lesion severity, prior response, pregnancy status, medication tolerance, and the possibility that a rash-like presentation is not acne at all. Any system that steps into that decision space must be able to sort straightforward cases from the ones that need human judgment.
Nolla’s current model still relies on physicians to close that gap. Two doctors review each prescription before release, which means the present version functions as a decision-support and workflow platform rather than a fully autonomous prescriber. Even so, the regulatory path now points toward something more radical: later phases in which the AI issues drugs independently. For a state to permit that progression places pressure on long-standing assumptions about who, or what, can safely initiate treatment.
The company’s public position also places liability squarely on Nolla. That matters because liability and oversight are not abstract legal details; they are the guardrails that determine whether patients have meaningful protection when software enters the prescribing chain. In a conventional clinic, a physician’s signature carries both clinical accountability and a built-in expectation of review. A software-driven prescribing model forces regulators to ask whether similar safeguards can be encoded into performance metrics, audit trails, and stop rules, or whether direct clinician oversight remains essential.
A facial scan is not the same as a dermatology examination.
That short distance between image capture and prescription hides much of the practical risk. Acne severity can be misread from photographs alone, especially when lighting, skin tone, camera quality, or overlapping facial conditions distort the image. Medical history also matters. Adult women may have acne tied to hormonal factors, some patients may have contraindications to specific medications, and others may need escalation beyond first-line therapy. If software is going to prescribe safely, it has to handle not just obvious lesions but also the exceptions that change the treatment plan.
The Utah authorization also arrives at a moment when health systems are already under strain from access problems and rising digital triage tools. AI is being inserted into more points of care, often first as an administrative filter, then as a recommendation engine, and finally as an autonomous actor. Prescribing is a higher-stakes step than most because medication errors can create direct harm, and the margin for false reassurance is small. Acne may be the opening indication, but the precedent could matter far beyond dermatology if regulators accept independent AI prescribing as a workable model.
For now, the details that matter most are the limits. Nolla is still in phase 1. Two licensed doctors review every prescription. The safeguards are tied to continued authorization. And the company says it remains responsible for the system’s performance. Those facts define a narrow but important starting point: artificial intelligence can now participate in prescribing in Utah, but not yet replace the clinician’s signature in practice. The next phase will reveal whether the state views that handoff as a controlled evolution or a line that should not be crossed.