Artificial intelligence just crossed a line it has never crossed before. For the first time in US history, an AI system has been given the legal authority to write a patient's very first prescription — no human doctor's signature required.
The milestone comes from Utah, where the state's Office of Artificial Intelligence Policy and its Division of Professional Licensing have co-signed a one-year agreement with healthcare startup Nolla Health. Through the company's Nolla Derm app, patients scan their face, answer a short health questionnaire, and the AI assesses acne severity and picks a personalized treatment plan. Starting this week, that AI can issue the actual prescription itself.
It is deliberately narrow. The pilot covers Utah residents aged 18 and older with mild-to-moderate acne, and the AI can only choose from eight approved topical treatments — no pills, no Accutane, no hormone therapies. The service costs $4.99 a month, and Nolla says the AI refers users to a human physician whenever it cannot confidently select a treatment.
The oversight scales up only as the AI proves itself. For the first 100 patients, two physicians review every AI-generated prescription before it is issued. The next 400 get physician review after the prescription goes out, and after that, monthly audits sample at least 10% of prescriptions plus every case involving an escalation or side effect.
This matters because healthcare access is a real bottleneck. Dermatologist waitlists stretch for weeks, and acne — while low-risk — is exactly the kind of routine, high-volume case that clogs specialist calendars. Nolla's pitch is that handling these cases with AI frees dermatologists to focus on complex, high-risk patients who genuinely need human expertise. Co-founder and CEO Luis Wenus called Nolla "the first ever end-to-end AI doctor" — a claim that rests on the fact that previous AI prescribing approvals in Utah covered only prescription refills, never a first prescription.
The honest caveat: this is one state, one condition, eight pre-approved treatments, and tight protocols. The safety assessment comes from the company, not from the pilot's data, which doesn't exist yet. But that is also the template. If the Utah experiment holds up under real physician audits, expect other states — and other low-acuity conditions — to follow. Autonomous medicine isn't arriving in one dramatic leap. It is arriving one bounded, boring, tightly supervised pilot at a time.
Comments
Post a Comment