No. 79
2026-10-06
the AI pulse — brewed, not raw
 AI CRITIQUE
Bir adam, yüzünde mavi bir ızgara katmanının vurguladığı yüz tanıma teknolojisini çağrıştırarak akıllı telefonuna bakıyor.

Utah launches first autonomous system to prescribe acne medication without doctor approval

Read the article    English (automatic translation)

Summary

The state of Utah has authorized Nolla Health's system to issue initial acne prescriptions without prior physician approval. The system selects from a pre-approved list of short-term topical treatments using facial scans and a structured questionnaire. Autonomy is being rolled out in phases, progressing through three stages from full physician review to weekly retrospective sampling.

Why it matters

This marks the first regulatory approval in the US for an AI to make an end-to-end initial prescribing decision. It sets a precedent for how states can test autonomous clinical judgment in a sandbox environment. The phased oversight model and strict safety measures (fixed drug list, mandatory human handoff) provide a replicable template for other states for narrow-scope, low-risk conditions. It also demonstrates that primary regulators for medical AI deployment are not just federal agencies, but also state-level AI policy offices.

Highlights

  • The system is locked to a short list of topical agents approved by physicians and must escalate any out-of-scope cases to a human doctor.
  • Nolla reports over 96% alignment with clinicians; however, this figure is self-reported and has not been independently verified.
  • Pricing starts at $4.99 per month. The app has reached over 175,000 downloads in 44 states; however, autonomous prescribing is limited to adults aged 18 and over in Utah.

Critical take

The 96% alignment rate is a company-provided metric that has not undergone independent audit. Since the pilot focuses solely on a narrow, low-risk condition (topical acne treatment), it is premature to generalize these results to broader autonomous prescribing. While the phased oversight design appears logical, the article provides no details on how adverse events or misdiagnoses will be tracked, reported, or corrected once the weekly sampling phase begins.

Bu kritik, yerel yapay zeka modeli Qwen3.8-27B tarafından yazılmıştır. Haberi ben seçtim; kritik metnini yayımlanmadan önce tek tek denetlemiyorum.This critique was written by the local AI model Qwen3.8-27B. I chose the story; I do not review each critique before it is published.
Back to the issue