Fact Check Team: Utah tests AI-powered prescribing without advance doctor approval
by MORGAN STEPHENS | Fact Check Team
Fri, October 16th 2026 at 12:00 PM
PLEASANTON, CA - DECEMBER 19: Volunteer Rebecca Cox (R) helps a man log perscription drug information as he registers his parents for the new Medicare drug prescription program during a Medicare enrollment event December 19, 2005 in Pleasanton, California. Open enrollment for the new program began November 15 and will continue through May 15, 2006. (Photo by Justin Sullivan/Getty Images)
WASHINGTON (TNND) — Utah is testing whether artificial intelligence can prescribe certain medications without a doctor approving every prescription in advance, marking a new experiment in how AI could be used in health care.
The state-supervised pilot involves Nolla Health, a health technology company whose AI-powered app, Nolla Derm, is authorized to prescribe certain topical acne medications to eligible adults in Utah.
But the program is limited in scope. It does not authorize artificial intelligence to prescribe any medication for any condition, and it includes physician oversight that changes as the pilot progresses.
How does Utah's AI prescribing pilot work?
Under the pilot's phased approach, physician review requirements depend on how far the program has progressed.
For the first 100 patients, two Utah-licensed physicians review every prescription before it is sent to a pharmacy.
For the next 400 patients, the AI system can issue prescriptions without advance physician approval. A physician reviews those prescriptions afterward.
After those initial phases, physicians review a sample of prescriptions, with additional oversight requirements established by the program.
The approach allows the AI system to issue certain prescriptions without a doctor signing off on each one beforehand, while retaining a role for human medical professionals.
Who is checking the AI?
Physician oversight remains part of the pilot, but the timing and scope of that review change as the program advances.
Initially, every prescription receives advance review by two physicians. In the next phase, prescriptions are reviewed after they are issued. Later, physicians monitor a sample of prescriptions rather than reviewing every prescription individually.
That distinction is important: Human oversight does not necessarily mean a physician approves every prescription before a patient receives it.
The program's safety will depend in part on how accurately the system evaluates patients, how effectively physicians identify potential problems, and how the program responds when errors occur.
Does this mean AI is replacing doctors?
Not entirely. Utah's pilot tests a specific use of AI in prescribing, rather than transferring all prescribing decisions to artificial intelligence.
The system is restricted to certain topical acne medications and eligible adults. Its phased review process also maintains physician involvement, although not every prescription requires advance approval once the program moves beyond its initial stage.
The pilot represents a test of how AI might handle a limited prescribing role under state-established conditions. Its broader implications for health care will depend on the results, including evidence about prescribing accuracy, patient safety and the effectiveness of ongoing monitoring.