Utah becomes first US state to let AI prescribe medication without doctor approval

Utah is on track to become the first US state where an AI system can examine patients and issue prescriptions without a doctor directly approving each decision. According to TechSpot, the program comes from Nolla Health, a healthcare startup whose app will diagnose and treat a narrow slice of dermatological complaints — and, once fully ramped up, do so with no clinician in the loop at the moment of prescribing.

How the service works

Customers pay $4.99 per month for the Nolla Health app and begin by scanning their face with a phone camera. After filling out a form covering medical history and personal details, the AI examines the skin, rates how oily it is, and produces a numerical severity score spanning the range from clear skin up to severe acne, TechSpot reports, citing Bloomberg.

That score is combined with the patient-supplied history to select a medication from a list of eight approved treatments. The scope is deliberately narrow: the program runs for one year and is restricted to people with mild-to-moderate acne.

Nolla Health CEO Luis Wenus told TechSpot that the downside risk is contained because all eight medications are topical rather than pills, with potential adverse effects limited to localized skin irritation or dryness.

A phased path to autonomy

The rollout is staged rather than immediate. The first 100 prescriptions must be reviewed by a physician before they reach the pharmacy. In the second phase, covering the next 400 patients, the AI submits prescriptions to pharmacies itself, with a doctor reviewing them retrospectively on a weekly basis. In the final phase, a physician reviews at least 10 percent of all issued prescriptions once a month.

The design amounts to a graduated experiment in autonomy: human involvement shrinks from per-prescription approval to periodic sampling, while the AI's authority expands to full prescriptive control.

A backdrop of AI health advice gone wrong

TechSpot frames the Utah move against a run of warnings about AI-generated health guidance. Last year, a 60-year-old man who followed ChatGPT's dietary advice replaced salt with sodium bromide and developed bromism, a psychiatric condition rarely seen since the 19th century.

On the platform side, Google quietly retired an AI-powered search feature called What People Suggest in March. It had surfaced crowdsourced advice from people with similar medical conditions, and its removal came a couple of months after the company pulled several AI Overviews when a report found users were being shown inaccurate health information.

Those cases involved general-purpose AI offering informal advice with no oversight structure. The Utah program is structurally different: a purpose-built medical system making treatment decisions inside a regulatory framework, with a defined drug list and mandatory audit sampling.

Why it matters

The immediate clinical exposure is small — acne, topical creams, a capped one-year program. The significance is regulatory. For the first time, a US state is treating an AI as the prescribing clinician rather than as a tool that assists one, and the phased-review structure gives regulators a template for how autonomy could be expanded condition by condition.

TechSpot notes the program could mark an early step toward AI standing in for doctors, or at least toward delegating less serious conditions to automated systems. Whether that path widens depends on what the audit data shows: if retrospective review turns up few problems, expect pressure to extend the model to other conditions and other states; if it does not, Utah becomes the reference case for what went wrong. Either way, the question of whether autonomous medical AI can be trusted with real patients now has its first live test in the United States, and the results will be watched far beyond dermatology.