Utah Turns an Acne App Into a Clinical AI On-Ramp

Nolla Health went live in Utah’s AI sandbox on October 5, 2026, writing first topical acne prescriptions for adults at $4.99 a month.

The same day, the Office of Artificial Intelligence Policy published a much larger package: master deals with University of Utah Health and Intermountain Health, a pelvic-floor therapy pilot, a chronic-drug refill pilot, and six outside auditors. Nolla is the only new company already running.

Dual Sign-Off Still Guards Every Acne Script

The New York company, operated by Magic Health, says it is the first in the country cleared by a state regulator for AI to issue an initial prescription, not just a refill. Utah residents 18 and older open the Nolla Derm app, verify a government photo ID, consent, answer a short intake, and take a five-angle face scan. In 10 to 15 minutes the system can return a treatment plan and, if the case fits, a prescription from a short list of creams and gels.

It does not write oral acne drugs or isotretinoin, and it does not treat severe acne. Prescriptions go only to Utah addresses. Pharmacists are told the order came from AI and are given a physician to call. Patients can message a licensed physician in the app at no extra cost, and those messages are answered by a person, not the model.

Luis Wenus, co-founder and chief executive, called the launch the first “end-to-end AI doctor.” Sean Geiger, who leads engineering at Nolla, put it in narrower terms: to his knowledge, this is the first time a U.S. regulator has allowed an AI to decide whether to start a patient on a prescription treatment.

Today, Nolla Health became the first organization in the U.S. (and possibly the world) to receive regulatory approval for an AI to issue initial prescriptions.

This makes Nolla the first ever actual end-to-end AI doctor. pic.twitter.com/xn6rvb8IyW

— Luis Wenus (@luiswenus) October 5, 2026

The signed file is tighter than that slogan. In the first stage, two physicians independently review and approve every AI-generated prescription before it reaches the patient. That stage covers the first 100 patients and runs at least four weeks. Moving on requires safety targets, including 95% agreement with physicians and zero serious adverse events, plus written approval from the state.

Nolla says licensed clinicians already agree with its treatment recommendations in 96% of real-world cases, a company figure separate from the 95% gate, and that the rest are small changes such as the strength of a topical. Dr. Zaid Fadul, the overseeing physician, said the model is not making freeform decisions.

There’s been a lot of noise about AI in healthcare, and a lot of it deserves scrutiny. What sets Nolla apart is the clinical rigor behind it. The AI isn’t making freeform decisions. It’s selecting from pre-approved protocols reviewed by physicians, and clinicians are built into every stage.

Dr. Zaid Fadul, overseeing physician, Nolla Health

On October 6, Wenus posted a clarification that the pilot starts with 100% physician oversight and that AI autonomy is earned only through demonstrated safety, with physicians still reviewing cases after the fact. The Division of Professional Licensing still lists a human physician as the prescriber of record on every order, under that physician’s NPI.

Two Hospital Systems Signed Master Agreements

OAIP’s October 5 newsroom notice listed five healthcare sandbox agreements, including two that do not start a product at all. University of Utah Health signed a master agreement on September 10, 2026. Intermountain Health, a Utah nonprofit system, signed its master in September 2026. Each sets standing terms so the office can approve later pilots one at a time, in a written addendum, instead of negotiating a full deal for every project.

On their own the masters authorize nothing. No pilot has been approved under either one. Each future addendum still has to name the exact rules being set aside. A complete request to move a phased pilot to its next phase that the office has not answered within 30 business days is treated as approved, a quiet accelerator sitting inside both hospital files.

Because the University is part of state government, its deal says it takes part only as a participant and has no role in reviewing its own eligibility or auditing its own pilots. Intermountain agreed that every tool in a pilot will follow its Responsible AI Use Policy, and that patient-safety and final-reporting duties continue even if funding for a pilot ends. Quarterly reports for each pilot are to be made public by the office.

The state also added a Healthcare pillar to its Pro-Human AI Initiative and named Tracy Gruber, commissioner of the Department of Health and Human Services, to the taskforce. Margaret Woolley Busse, commerce commissioner and taskforce co-chair, said the office is trying to build a model where AI can improve outcomes while oversight stays in place. Data from the pilots is supposed to go to the Legislature and the public before anyone writes a permanent rule.

WHAT WE KNOW

• Nolla’s clock: The acne pilot is listed as active from October 5, 2026, through October 5, 2027.
• Hospital paper: Both master agreements are signed, and neither has an approved addendum yet.
• No endorsement: Acceptance into the sandbox is not a state endorsement of any product.

WHAT IS UNCONFIRMED

• Next phases: Whether Nolla will receive written approval to drop dual sign-off after the first 100 patients.
• Start dates: When August AI and Expect Fitness will open their demonstration periods.
• Hospital tools: Which clinical AI, if any, University of Utah Health or Intermountain will file first.

Zach Boyd, director of OAIP, said technical innovation will never supersede patient safety, and that independent evaluators are there so every claim from a developer is checked. The office cannot grant relief alone. The head of the relevant agency, most often the Division of Professional Licensing, has to sign too.

Pelvic Floor Therapy and Chronic Refills Wait

The acne app is the one patients can open now. The public register of AI pilots still lists the other two new company tests as approved but not started. Both were signed on October 2, 2026, with the Division of Professional Licensing.

THREE COMPANY PILOTS ON THE STATE REGISTER

August AI, operated by August Labs, may not write a new prescription, change a dose, or substitute a drug. Identity is verified and the pharmacy record is checked before any refill. No more than three AI refills in a row are allowed at the start, after which the patient must see a human provider. After the first 250 refills, the next 1,000 are reviewed after they go out. Later, providers spot-check at least 5% of refills in every medication class. Each new phase needs written approval. Suicidal thinking, a new side effect, a dangerous interaction, or missing lab monitoring stops the refill.

Expect Fitness scores answers to standard questionnaires and drafts an exercise and care plan. Higher-risk patients always get full therapist review, and any adverse event returns the platform to full review. Unexplained bleeding, fever with pelvic or urinary symptoms, or chest pain stops the process and sends the patient to urgent care. The office gets a live dashboard of adverse events, worsening symptoms, and escalations, with serious events due by the end of the next business day.

Those two files sit in the same legal family as Nolla, and they are the part of the October 5 announcement that the acne headline buried. One is women’s health. One is chronic disease. Both are built to shed live clinician review if the data, and the office, allow it.

Utah’s Medical Board Was Blindsided Once Already

This expansion did not arrive on a clean slate. An earlier refill pilot with Doctronic, which went public on January 6, 2026, after a 12-month clock that began in October 2025, produced a fight with the people who license physicians in the state.

HOW THE SANDBOX GOT TO FIRST PRESCRIPTIONS

1. May 2024: The Legislature creates the Office of Artificial Intelligence Policy under the Artificial Intelligence Policy Act, with power to adjust how a Utah rule applies to a specific AI product for a limited test.
2. January 6, 2026: Doctronic’s refill pilot is launched in public. The system may renew existing prescriptions on a fixed list; it is not allowed to start a new drug.
3. April 20, 2026: Members of the Utah Medical Licensing Board send a letter demanding the Doctronic pilot be suspended at once, saying they were not consulted before it went live.
4. October 5, 2026: OAIP adds a healthcare pillar, six outside evaluators, two hospital masters, and the Nolla, August AI, and Expect Fitness deals. Nolla Derm goes live the same day.

Alan Smith, chair of the Utah Medical Licensing Board, said the Doctronic agreement was already approved and running before members even knew the refill pilot existed. “We were kind of blindsided and quite concerned about it.” Board members wrote that professionals with medical backgrounds should review all proposals before they launch, and that they must not allow AI or other financial motives to override that duty. The office declined to suspend the pilot, noting that in its then-current phase a licensed physician reviewed every AI-generated renewal before it reached a pharmacy.

The Utah Academy of Family Physicians said it had “significant concerns” and called the move an apparent willingness to go forward without the guardrails it wanted. Robert Steinbrook, health research group director at Public Citizen, called the refill program “a dangerous first step toward more autonomous medical practice.”

Nolla’s file looks written to avoid a repeat of that ambush. The company says it met with the medical board before launch and ran the protocol through practicing physicians, dermatologists, and state officials over several rounds. The FAQ now describes a Medical Board AI Working Group that meets with OAIP on proposals. Legion Health, a psychiatry refill pilot signed March 19, 2026, is still listed as not started, another reminder that a signature is not a launch.

What Happens After the First 100 Patients?

Nolla’s own staging plan, which still needs the office’s written OK at each step, is the clearest map of where this is meant to go. Stage two, up to 500 patients and at least eight weeks, lets Nolla Derm issue prescriptions directly, with a physician reviewing every case after the fact, at least weekly. Stage three has physicians review a sample of at least 10% of prescriptions every month, plus every case with an escalation or a side effect.

That is the second-order move. Dual sign-off is the opening condition, not the end state. The state register says later stages move to review after the fact, then to sampling. It also says each of those steps requires written approval from the office. The product can get more autonomous. It cannot do so on a calendar alone.

WHEN NOLLA DERM MUST STOP AND HAND OFF

• Severe acne: The pilot is limited to mild to moderate disease; severe cases are screened out of the AI path.
• Pregnancy: Pregnancy, plans to become pregnant, or breastfeeding stops the process.
• Immune risk: A weakened immune system, or a past reaction to a listed medication, stops the process.
• Out of bounds: Oral medication and isotretinoin are excluded, and other conditions outside the protocol go to a licensed physician or in-person care.

The prescribing flow has no free-text box, so the system cannot invent a treatment off the list. Identity and age are checked with a government photo ID. Adverse events go to the office within 24 hours, and monthly reports include how often the AI and reviewing physicians agreed. Terms of service may not limit the company’s liability for harm, and professional liability insurance is required. Patient data may not be sold.

The same stage logic appears in the unstarted files. August AI moves from 250 live reviews to 1,000 after-the-fact reviews to a 5% spot check. Expect Fitness starts with a therapist on every plan and only later tests whether the lowest-risk plans can go out with spot checks. Across all three, the destination is less pre-review, on evidence the office accepts.

Six Outside Groups Now Audit the Pilots

The other half of the October 5 package is a standing audit bench. OAIP signed six independent evaluators to check technical attestations, audit the accuracy of AI decision reporting, and run clinical evaluations against medical standards. The six are the Coalition for Health AI, Clarion AI Partners, Glacis Technologies, mpathic AI, Vega Health, and Stanford’s Clinical Excellence Research Center.

That network is the state’s answer to the charge that it waved refill tools through without enough clinical eyes. New proposals, the office says, will also go through these evaluators. Boyd has described the method as market-based: let the market and the state learn together, with a validation layer that can grow as the tools grow.

Glacis’s piece of the stack is concrete. A memorandum of understanding makes a cryptographically signed receipt mandatory for every covered AI inference, or for an associated safety or governance check, using an open evidence standard called OVERT. Joe Braidwood, founder and chief executive of Glacis, said clinicians and patients need to be able to verify that safeguards are actually running, and that Utah’s deal puts a receipt behind every AI decision. OAIP still decides which vendors, systems, and events are covered.

The legal power underneath all of this is limited, and it is written down. An agreement can adjust how a specified Utah rule applies to one participant for a set period. It is not a general license to practice medicine. The office’s statute on regulatory mitigation requires the deal to state the limits on scope, the safeguards, the disclosures owed to consumers, and the reporting the office will audit. An initial demonstration period may not exceed 12 months. A participant may request a 12-month extension, and the office may grant up to two. The office may end an agreement at any time. Ending it restores the old rule; it does not leave a leftover permission slip.

Relief reaches only the sections named in each schedule. For Nolla, the Division agreed not to bring an unlicensed-conduct or telehealth enforcement action where the AI issues a topical acne prescription under the protocol, and the same protection covers the physicians who rely on the system for that purpose. Federal requirements are untouched. Patients keep ordinary malpractice remedies. Companies cannot broadly disclaim medical harm in their user terms, a clause the office added after earlier pilots drew fire on that point.

Eleven Counties Have No Dermatologist at All

Nolla’s pitch is access, priced to undercut an office visit. The Utah version of the app is $4.99 a month; medication is extra. The company raised a $4.5 million General Catalyst round in September 2025, with SNÖ Ventures and Commure founder Diede van Lamoen also in the round, and it says the app has already been downloaded by more than 175,000 people across 44 states under clinician guidance. Daily skin scans track progress, and plans are reviewed after every 28 scans, about every one to two months.

THE ACCESS GAP NOLLA CITES IN UTAH

• Who has acne: About 100,000 Utahns have active acne at a given time, the company says.
• Who sees a dermatologist: Only 10% ever do, in Nolla’s figures.
• The wait: The company puts the average appointment wait in Utah at 61 days, nearly double the national median.
• The map: Eleven counties have no dermatologist, and a visit can cost up to $300.

Wenus has said dermatologists in a high-melanoma state are already stretched by routine cases and still cannot cover the acne patients who need care. The long-term plan, which he has also posted, is to start in dermatology and expand into women’s health and other specialties. That is the same direction the sandbox itself just took, with Expect Fitness queued behind Nolla on pelvic-floor care.

The live product is still a supervised acne protocol with two Utah licenses on every script. The files signed around it are a template for first prescriptions, chronic refills, women’s health therapy, and whatever University of Utah Health and Intermountain put in an addendum. Nolla’s demonstration period ends on October 5, 2027, unless the office extends it. August AI and Expect Fitness are still waiting to start.

Disclaimer: This article is news reporting on a state-supervised test of clinical software, and it is for information only. It is not medical advice, a diagnosis, or a recommendation to start, stop, or refill any medication, and it is not legal advice about licensing or malpractice. Anyone considering using one of these apps, or changing a treatment plan, should talk with a licensed physician, pharmacist, or physical therapist who can review their own history. Figures, start dates, and oversight stages reflect the Office of Artificial Intelligence Policy register, company statements, and related primary documents available on October 11, 2026, and those terms can change when the office issues written approvals or ends an agreement.