Utah Just Approved AI to Diagnose Acne and Write New Prescriptions. Here’s What It Means for the Future of Healthcare

An artificial intelligence system can now assess acne from smartphone photos and issue first-time topical prescriptions under a limited Utah pilot. The move could reshape access to routine care, but it also raises important questions about medical oversight, patient safety, and accountability.

Imagine opening an app, taking five photos of your face, answering a few questions, and receiving an acne treatment plan that includes a new prescription, without booking an in-person appointment.

This is no longer just a vision of the future. As of October 2026, Utah has authorized Nolla Health to test an AI-powered acne treatment system under a state regulatory pilot. Unlike earlier AI prescription programs focused on renewing medications already prescribed by a physician, this initiative allows the system to assess a patient's initial presentation and issue a new prescription for certain topical acne medications within a tightly defined scope.

The development is significant because it moves artificial intelligence beyond answering health questions or summarizing medical records. It places AI directly inside a clinical decision-making process that can result in medication being prescribed.

For patients, this could mean faster access to treatment and fewer barriers to receiving care. For healthcare professionals, it raises a more complex question: how can AI take on specific clinical responsibilities without compromising patient safety or professional accountability?

Dr. Zaid Fadul, a board-certified family physician and Chief Medical Advisor for Nolla Health, has discussed the pilot and its implications for the future of healthcare. Fadul also serves on the company's Medical Advisory Board and holds a small ownership stake, a relevant disclosure when evaluating his perspective on the initiative.

What Makes Utah's AI Prescription Pilot Different?

AI is already used in healthcare for tasks such as analyzing medical images, identifying patterns in clinical data, assisting with documentation, and supporting clinical decisions. However, there is a fundamental difference between an AI tool that provides information to a physician and a system authorized to initiate treatment.

An AI assistant might help a doctor evaluate a patient's symptoms or suggest possible treatment options. A prescription-refill system, by contrast, may be authorized to renew a medication that a clinician previously prescribed, provided the patient meets defined criteria.

The Utah pilot takes another step forward.

Under the agreement with Nolla Health, the AI system can assess acne from photographs and issue first-time prescriptions for approved topical treatments for eligible adults with mild to moderate acne. The authorization does not extend to every medication, every patient, or every skin condition. Oral medications, isotretinoin, and severe acne are outside its permitted scope.

That distinction matters. Starting a new treatment requires evaluating a patient's current condition, determining whether a treatment is appropriate, and identifying circumstances in which medication could be unsafe.

The pilot is therefore not a general authorization for AI to practice medicine independently. It is a limited experiment in allowing an AI system to perform a defined clinical task under regulatory conditions.

Why Acne Is a Logical Starting Point for AI Healthcare

Acne is common, can persist into adulthood, and can affect both physical appearance and emotional well-being. Yet access to dermatology services is not always straightforward.

Patients may face long appointment waits, travel requirements, insurance limitations, or out-of-pocket consultation costs. Others try over-the-counter products for months without receiving a structured assessment or evidence-based treatment plan.

Research published in PLOS ONE in 2023 examined U.S. National Ambulatory Medical Care Survey data from 2002 through 2016. It found that, among adults older than 20, acne accounted for 10.1% of dermatology visits by women compared with 4.1% of visits by men. The findings demonstrate that acne remains a meaningful part of adult dermatology care, rather than being exclusively an adolescent concern.

Acne is also a condition that can often be evaluated through a combination of visual examination and patient history. Its visible features make photographic assessment a potentially useful component of a digital care pathway.

However, photographs alone cannot reveal everything about a patient's health. A clinician may need to consider medical history, allergies, pregnancy, previous medication reactions, other conditions, and whether the skin problem is actually acne.

This is why the choice of acne as an initial use case is important. It offers a relatively narrow clinical problem for testing AI-assisted assessment and treatment, while still requiring meaningful safeguards.

The goal is not to demonstrate that AI can replace dermatology. It is to determine whether a carefully restricted system can safely manage selected, lower-risk cases and identify when a patient needs a different level of care.

How the AI Acne Diagnosis Process Works

The Nolla Derm model uses photographs of a patient's face to assess acne. The company's public description presents a digital process in which patients provide images, receive a treatment plan, and track changes in their skin over time.

A typical pathway involves several stages.

1. Patient identification and eligibility

The system first needs to establish that the patient is eligible for the pilot. Utah's agreement requires age and identity verification, and prescriptions are restricted to Utah addresses.

2. Photographic skin assessment

The patient submits facial photographs through the app. The AI evaluates visible acne features to support its assessment of severity and determine whether the case falls within the permitted treatment pathway.

3. Treatment selection

For eligible patients, the system selects from the limited list of approved topical medications. It cannot freely choose any medication available in dermatology, nor can it prescribe oral acne medications or isotretinoin under this agreement.

4. Prescription and pharmacy fulfillment

The prescription is sent to a Utah pharmacist under the pilot's protocol. Pharmacists are informed that the prescription originated through AI and are provided with a physician contact for questions.

5. Monitoring and follow-up

The platform also offers ongoing skin tracking, allowing patients to monitor changes over time. Persistent symptoms, worsening acne, adverse reactions, or circumstances outside the system's permitted scope may require a human clinician to intervene.

The critical point is that automated assessment does not eliminate the need for clinical judgment. The system must operate within a defined pathway and stop when a patient falls outside it.

Utah's AI Regulatory Sandbox: How the State Made It Possible

The legal foundation is Utah's Artificial Intelligence Policy Act, established through Senate Bill 149 in 2024, and the state's regulatory learning laboratory program.

The legislation created the Office of Artificial Intelligence Policy (OAIP), housed within the Utah Department of Commerce. Its role includes examining how AI technologies operate in regulated industries, protecting the public, and exploring whether existing rules need to change as technology develops.

The regulatory sandbox provides a way to test a specific technology under a written agreement that temporarily modifies the application of certain regulations. Companies must work with the relevant regulatory agency, which participates in approving the agreement.

This does not mean Utah has eliminated medical regulation or created a blanket exemption for AI healthcare companies. The relief is narrow, time-limited, and tied to the approved product and activities.

Nolla Health's pilot runs from October 5, 2026, through October 5, 2027, with the possibility of extensions under the state's framework. The state can also terminate an agreement, and participation does not mean Utah has endorsed or certified the product.

This distinction is important for other healthcare companies considering AI adoption. Utah is testing whether specific regulatory requirements can be adapted while preserving safeguards, rather than allowing companies to bypass the law altogether.

What Safety Limits Prevent the AI From Going Too Far?

The pilot's restrictions are among its most important features. It is designed to serve a defined patient population rather than attempt to manage every possible acne case.

According to the state's published agreement, the system must stop the automated process when patients fall outside its eligibility criteria. Exclusions include:

• People younger than 18.
• Patients with severe acne.
• People who are pregnant, planning pregnancy, or breastfeeding.
• Patients with weakened immune systems.
• Patients with a previous reaction to one of the medications covered by the protocol.

The AI is also prohibited from prescribing oral acne medications or isotretinoin, commonly known as Accutane. Its prescribing authority is restricted to the approved list of topical treatments.

These restrictions help define the boundaries of the system. An AI designed to manage uncomplicated acne should not attempt to treat a patient whose condition may require a specialist, a different diagnosis, or more extensive medical evaluation.

The pilot also includes physician oversight, reporting requirements, and a requirement for professional liability insurance. During the initial stage, two Utah-licensed physicians independently review and approve every AI-generated prescription before it reaches the patient. Later stages may reduce the level of advance review, but progression requires meeting defined safety targets and obtaining written state approval.

These measures are intended to make the pilot measurable and accountable. Their effectiveness, however, will ultimately depend on actual clinical outcomes, the quality of the AI's assessments, and how consistently the safeguards are followed.

How Accurate Is AI at Diagnosing Acne?

AI-assisted skin assessment is not entirely new. Researchers have studied whether computer vision can identify acne lesions and grade their severity using smartphone photographs.

A study published in Experimental Dermatology in 2019 examined an AI algorithm trained using 5,972 smartphone images from 1,072 acne patients. Three dermatologists assessed the patients' acne severity, providing a clinical reference for the algorithm.

The study reported that the final version of the algorithm achieved 68% agreement-level performance on its acne severity grading measure, comparable to the dermatologist assessments used in the study. It also reported different precision and recall results for identifying inflammatory lesions, non-inflammatory lesions, and post-inflammatory hyperpigmentation.

These results are encouraging, but they require careful interpretation.

The 68% figure is not proof that every AI acne diagnosis is 68% accurate, nor does it establish the performance of Nolla Health's particular system. The research evaluated a different algorithm under specific study conditions.

There is also a difference between recognizing visible acne and deciding whether a particular medication is safe for a particular person. A photograph-based system might grade the visible severity of a skin condition reasonably well while still needing patient history and clinical safeguards to make appropriate treatment decisions.

To evaluate Nolla Derm independently, researchers and regulators would ideally examine its performance across different skin tones, image quality, acne presentations, and patient populations. They would also need data on inappropriate prescriptions, missed warning signs, adverse events, and treatment outcomes.

The relevant question is not simply whether AI can recognize acne. It is whether the complete care pathway can consistently identify suitable patients, select appropriate treatments, and escalate cases that require human judgment.

Who Is Legally Responsible If the AI Gets It Wrong?

This is one of the most consequential questions raised by AI prescribing.

An AI system does not take personal legal responsibility in the same way a physician or company can. When a patient is harmed, responsibility may depend on the circumstances, applicable law, professional duties, the company's conduct, and the role of any clinicians involved.

Utah's pilot agreement addresses several aspects of accountability. Nolla Health must maintain professional liability insurance, report adverse events to the state within 24 hours, provide monthly information about the program, and comply with the agreement's restrictions. The agreement also states that the company's terms of service may not limit its liability for harm.

The pilot's regulatory protection is not a blanket immunity from legal claims. Utah's sandbox framework explicitly preserves obligations outside the provisions that have been modified.

That means the question of responsibility cannot be reduced to saying that the physician is always liable or that the technology company is automatically responsible for every error. The facts of a particular case matter.

For healthcare organizations considering similar systems, this makes governance essential. They need to establish who approves clinical protocols, who monitors system performance, who responds to adverse events, how patients can reach a human clinician, and what insurance applies if an AI-generated decision causes harm.

Transparency is equally important. Patients should understand when AI is involved in their care, what the system can and cannot do, and how to obtain further medical attention when needed.

Does AI Prescribing Mean Doctors Are Being Replaced?

Not necessarily. A more useful way to view this development is as a test of whether some routine clinical tasks can be automated safely while physicians remain responsible for oversight, escalation, and more complex care.

Healthcare systems face pressure from rising demand, administrative work, uneven access to specialists, and workforce constraints. AI may help address some of these problems by managing narrowly defined tasks that otherwise require an appointment or a clinician's time.

But efficiency alone is not enough to justify automating healthcare.

A successful system must demonstrate that it can deliver appropriate care, protect patient information, recognize its limitations, and direct patients to human professionals when necessary.

For physicians, AI could eventually reduce the time spent on repetitive assessments and routine cases. That could create more capacity for patients with complicated conditions who need detailed evaluation and individualized medical judgment.

For patients, the potential benefits include convenience, lower barriers to treatment, and easier follow-up. The risks include incorrect assessments, unsuitable prescriptions, missed diagnoses, and overconfidence in automated recommendations.

The outcome will depend on how carefully these systems are designed, monitored, and integrated into existing healthcare services.

Why the $4.99 Monthly Price Matters

Nolla Health advertises its Utah pilot from $4.99 per month, with medication costs billed separately. The company also describes digital skin tracking and access to a physician through its platform.

The low advertised membership price is significant because the cost of care can discourage patients from seeking help for conditions they perceive as non-urgent. A digital service could make initial treatment more accessible, particularly for people who face financial or scheduling barriers.

However, the subscription price should not be mistaken for the total cost of treatment. Patients should consider medication charges, the services included in the membership, and whether additional medical care might be needed.

Nor does a lower price automatically establish better clinical outcomes. The real measure of success is whether the service helps eligible patients receive safe, appropriate treatment at a cost they can manage.

The broader opportunity is to make healthcare more convenient without reducing it to a technology transaction.

Could Other States Follow Utah's Lead?

Utah's approach could influence discussions in other states about how to regulate AI in healthcare. However, the pilot does not automatically authorize AI prescribing elsewhere in the United States.

Other jurisdictions have their own medical licensing rules, prescribing requirements, telehealth regulations, and approaches to AI oversight. A company operating across state lines would need to examine the requirements that apply in each jurisdiction.

The Utah experiment may provide useful evidence about how to structure clinical AI pilots, including eligibility criteria, staged physician oversight, adverse-event reporting, liability insurance, and independent evaluation.

The results could also inform decisions about which types of clinical tasks are suitable for automation and which require direct human involvement.

For now, the authorization is narrow: adult acne treatment, approved topical medications, defined eligibility criteria, and a limited regulatory period.

Whether the model expands will depend on what the pilot demonstrates about patient safety, clinical performance, and the practical value of the service.

The Future of Healthcare May Be More Automated, but It Must Remain Accountable

Utah's AI acne prescribing pilot represents an important shift in the discussion about artificial intelligence in medicine. AI is moving beyond providing information to participating in a process that can initiate treatment.

That does not establish that AI is ready to replace physicians. Instead, it creates an opportunity to test a carefully bounded model of digital care in which automation, regulation, and clinical oversight must work together.

If the pilot demonstrates reliable outcomes and effective safeguards, it could offer lessons for other routine healthcare applications. If it exposes weaknesses, those findings will be equally important in determining where automated prescribing should stop.

The central question is not whether AI can perform a medical task. It is whether it can perform that task safely, transparently, and with clear accountability when something goes wrong.

That is the standard by which the next generation of AI-enabled healthcare should be judged.

About Dr. Zaid Fadul

Dr. Zaid Fadul is a board-certified family physician with additional board certification in addiction medicine, Harvard leadership training, and experience as a U.S. Air Force flight surgeon. He founded Bespoke Concierge MD, a physician-led primary and preventive care service, and serves as Chief Medical Advisor to Nolla Health.

Disclosure: Dr. Fadul serves on Nolla Health's Medical Advisory Board, holds a small ownership stake in the company, and has a physician group that oversees the pilot. Readers should consider this relationship when evaluating his commentary on the program.

Learn more about membership options at Bespoke Concierge MD.

References and Further Reading

1. Utah Department of Commerce, Office of Artificial Intelligence Policy. Authorized AI Pilots: Nolla Health acne treatment agreement. Official source for the pilot's scope, restrictions, oversight, and reporting requirements.
2. Utah Department of Commerce. AI Policy and Regulatory Sandbox. Information about Utah's AI regulatory framework.
3. Utah Legislature. Senate Bill 149 (2024): Artificial Intelligence Policy Act. Legislative background for the state's AI policy and regulatory learning laboratory.
4. Utah Department of Commerce. AI Frequently Asked Questions. Explanation of regulatory mitigation, clinical AI pilots, and safeguards.
5. Chang J, Nock MR, Cohen JM, Bunick CG. “Acne accounts for an almost 2.5-fold higher proportion of dermatology visits among adult females compared to adult males in the United States.” PLOS ONE, September 21, 2023. Read the peer-reviewed study.
6. Seité S, et al. “Development and accuracy of an artificial intelligence algorithm for acne grading from smartphone photographs.” Experimental Dermatology, 2019. Read the study abstract and findings.
7. Nolla Health. Nolla Derm AI prescriptions: Utah pilot information. Company-published information about the service, pricing, and patient experience.
8. Nolla Health. Announcement of the AI-powered prescription pilot. Company announcement and description of its proposed staged physician oversight.

Editorial note: This article distinguishes between Utah's official regulatory authorization, published scientific research on acne-assessment AI, and claims made by Nolla Health about its own product. Independent outcome data for the Nolla Derm pilot should be reviewed as they become available.

‍