Utah has become the first state to create a regulated pathway for an artificial‑intelligence system to generate an initial prescription without a doctor’s pre‑approval. The Utah Office of Artificial Intelligence Policy signed an agreement with Nolla Health on September 22, and the company announced on October 5 that residents age 18 and older could begin enrolling.
How the pilot works
During the first stage, every AI‑generated prescription must be reviewed by two Utah‑licensed physicians before the medication is sent to a pharmacy. The agreement allows the AI to move to direct submission only after the pilot has run for at least four weeks, treated a minimum of 100 patients, achieved at least a 95 % agreement rate with physicians, recorded no serious adverse events, and missed no required safety stops. Utah regulators will decide whether to advance the program.
What the system does
Nolla’s platform is not a general‑purpose chatbot. It uses a structured medical intake, five‑angle facial images, and predefined clinical protocols to assess adults with mild to moderate acne. Treatment is limited to topical products such as tretinoin, spironolactone (compounded), niacinamide, adapalene, clindamycin with benzoyl peroxide, benzoyl peroxide and azelaic acid. Oral drugs, systemic hormonal therapy, isotretinoin and severe‑acne regimens are excluded.
The AI includes hard stops for pregnancy or plans to become pregnant, breastfeeding, weakened immune systems, prior medication reactions, severe acne, or poor‑quality images. Patients may also request a physician at any point.
Future stages
If the pilot meets its safety benchmarks, Stage 2 would let the AI submit prescriptions directly while physicians review each case at least weekly. Stage 3, triggered after 750 cumulative patients, would reduce routine physician review to a monthly sample of at least 10 % of enrollments, though doctors would still evaluate every adverse event and escalation.
Safety and oversight
Utah regulators have added independent third‑party evaluators to stress‑test the AI’s guardrails, inspect decision logs and assess patient outcomes. The Utah Medical Licensing Board, which previously urged suspension of an AI refill pilot, emphasized that prescription refills traditionally require physician authorization. The American Medical Association also cautioned that AI should support, not replace, physician judgment.
Nevertheless, Nolla reports that licensed clinicians agreed with its AI‑generated treatment recommendations in about 96 % of the last 1,000 treatments, with disagreements limited to minor strength adjustments rather than different therapeutic approaches. While these figures come from the company’s own data, the state‑mandated external audit will provide an independent check.
Why it matters
Medication errors remain a significant public‑health concern. The Centers for Disease Control and Prevention estimates that adverse drug events send more than 1.5 million Americans to emergency departments each year and cause nearly 500 000 hospitalizations. Automated cross‑checking of structured information, which computers excel at, could help reduce such errors.
A 2005 outpatient study of Boston primary‑care practices found prescribing errors in 7.6 % of 1 879 prescriptions, with about 3 % posing a risk of patient injury. Researchers estimated that advanced dose and frequency checks could have prevented 95 % of those potential adverse events. Although the study is dated, it underscores the long‑standing interest in using technology to improve medication safety.
Implications for other states
Texas has already enacted a legal framework that could support a similar experiment. House Bill 149, effective January 1, created an AI Regulatory Sandbox Program that includes health care, allowing approved participants to test AI systems for up to 36 months without traditional licensing, subject to state oversight. Senate Bill 1188, effective September 1, 2025, requires licensed practitioners to review AI‑generated records and disclose AI use to patients, preserving human control over critical decisions.
Utah’s pioneering pilot therefore offers a model that other states may watch closely as they balance innovation with patient safety and physician oversight.
Original reporting: The Dallas Express — read the source article.
