Authorized AI Pilots

What the pilot does

Expect Fitness runs an online pelvic floor physical therapy program for adult women in Utah with conditions such as urinary incontinence, pelvic pain, and postpartum pelvic floor problems. Its system scores a patient's answers to standard clinical questionnaires and drafts an exercise and care plan. A licensed physical therapist reviews and approves every plan before the patient sees it, and the pilot tests whether the lowest risk plans can later go out with spot checks instead.

Relief granted

The Division of Professional Licensing agreed not to bring an enforcement action for unlicensed or unprofessional conduct, or under the state telehealth requirements, where care plans are generated by the AI and delivered under the levels of physical therapist oversight set out in the agreement. The agreement also treats the platform as meeting Utah's AI disclosure law and the rule that a physical therapist manage each patient's care, as long as it follows the agreed protocols. The relief runs to the sections named in Schedule A and expressly reaches nothing else Expect Fitness offers.

What the company agreed to

• A licensed physical therapist reviews every plan in the first phase and can change or reject any of them. Moving to a later phase requires written approval from this office.
• Plans for higher risk patients always get full therapist review, and any adverse event returns the platform to full review.
• Warning signs such as unexplained bleeding, fever with pelvic or urinary symptoms, or chest pain stop the process and send the patient to urgent care.
• Patients are told AI is part of their care and must acknowledge it before using the platform.
• This office has a live dashboard of adverse events, worsening symptoms, and escalations, and serious events are reported by the end of the next business day.
• Professional liability insurance, and terms of service that may not limit the company's liability for harm.

What the pilot does

August AI, operated by August Labs, Inc., offers an AI health companion app. The pilot lets its system authorize routine 30, 60, and 90 day refills of maintenance medication a Utah adult is already taking on an established treatment plan, from a fixed list of non controlled medications set out in the agreement. It does not write a new prescription, change a dose, or substitute one medication for another.

Relief granted

The Division of Professional Licensing agreed not to bring an enforcement action for unlicensed or unprofessional conduct, or under the state telehealth requirements, where the AI authorizes a routine refill of an existing prescription under the agreed protocol. The relief runs to the sections named in Schedule A and expressly reaches nothing else August AI offers.

What the company agreed to

• A Utah licensed provider reviews the first 250 refills before they reach the patient, and reviews each of the next 1,000 afterward. Each new phase requires written approval from this office.
• After that, providers spot check at least 5 percent of refills in every medication class.
• Identity verification, and a check of the patient's pharmacy record to confirm the medication is still current, before any refill.
• No more than three AI refills in a row for any medication at the start, after which the patient must see a human provider.
• Any sign of suicidal thinking, a new side effect, a dangerous interaction, or missing lab monitoring stops the refill and sends it to a licensed provider.
• Pharmacists are told the refill came from AI, adverse events are reported to this office within 24 hours, and patient data is not sold or used for advertising.

What the pilot does

Nolla Health, operated by Magic Health, Inc. and Magic Health Services, LLC, makes an app that assesses acne from photos a patient takes of their face. The pilot lets its system issue first time prescriptions and refills for topical acne treatment to Utah adults with mild to moderate acne, from a short list of creams and gels set out in the agreement. It does not prescribe oral medication or isotretinoin, and it does not treat severe acne.

Relief granted

The Division of Professional Licensing agreed not to bring an enforcement action for unlicensed or unprofessional conduct, or under the state telehealth requirements, where the AI issues a topical acne prescription and sends it to a Utah pharmacist under the agreed protocol. The same protection covers the physicians working with Nolla Health who rely on the system for that purpose. The relief runs to the sections named in Schedule A and expressly reaches nothing else Nolla Health offers.

What the company agreed to

• Two Utah licensed physicians review every prescription before it is sent in the first stage. Later stages move to review after the fact, then to sampling, and each requires written approval from this office.
• Severe acne, pregnancy or plans to become pregnant, breastfeeding, a weakened immune system, or a past reaction to a listed medication stops the process.
• Identity and age verified with a government photo ID, and prescriptions sent only to Utah addresses.
• Pharmacists are told the prescription came from AI and given a physician to contact with questions.
• Adverse events reported to this office within 24 hours, and monthly reporting that includes how often the AI and reviewing physicians agreed.
• Professional liability insurance, no sale or transfer of patient data, and terms of service that may not limit the company's liability for harm.

What the agreement does

This is a master agreement, not a pilot. It sets standing terms under which University of Utah Health can propose AI pilots across its system and this office can approve them one at a time, each in a written pilot addendum, instead of negotiating a full agreement for every project. On its own it authorizes nothing.

Because the University is part of state government, the agreement says it takes part only as a participant, and that it has no role in this office's review of its eligibility or its audit of its own pilots.

Relief granted

None yet. Any relief must be written into a pilot addendum, and each addendum must include a schedule stating exactly which rules are waived or modified, any cap on penalties, the time allowed to fix a problem, and any safe harbor. If a problem arises under an approved pilot, the University gets written notice and at least 30 days to fix it before enforcement, unless there is gross negligence or willful misconduct.

Where it stands

No pilot has been approved under this agreement yet. Each one will be added to this page when it is.

What the organization agreed to

• No pilot begins until this office approves a written addendum for it.
• A phased pilot moves to its next phase only with this office's written approval. A complete request the office has not answered within 30 business days is treated as approved.
• Monthly, quarterly, and annual reports for each pilot, with quarterly reports made public by this office.
• No selling user data and no undisclosed targeted advertising.
• No advertising that suggests the state endorses the University or its technology.
• The safeguards, testing, and oversight set out in each addendum, under the same cybersecurity protections required of every participant.

What the agreement does

This is a master agreement, not a pilot. It sets standing terms under which Intermountain Health, a Utah nonprofit health system, can propose AI pilots and this office can approve them one at a time, each in a written pilot addendum, instead of negotiating a full agreement for every project. It covers Intermountain's hospitals, clinics, and subsidiaries. On its own it authorizes nothing.

Relief granted

None yet. Any relief must be written into a pilot addendum, and each addendum must include a schedule stating exactly which rules are waived or modified, any cap on penalties, the time allowed to fix a problem, and any safe harbor. The state keeps full authority over anything outside an approved addendum. If a problem arises under an approved pilot, Intermountain gets written notice and at least 30 days to fix it before enforcement, unless there is gross negligence or willful misconduct.

Where it stands

No pilot has been approved under this agreement yet. Each one will be added to this page when it is.

What the organization agreed to

• No pilot begins until this office approves a written addendum for it.
• Every AI tool used in a pilot follows Intermountain's own Responsible AI Use Policy.
• A plan for each pilot to manage its identified risks, including insurance for any risk to patients.
• A phased pilot moves to its next phase only with this office's written approval. A complete request the office has not answered within 30 business days is treated as approved.
• Monthly, quarterly, and annual reports for each pilot, with quarterly reports made public by this office.
• No selling user data and no undisclosed targeted advertising. If funding for a pilot ends, its patient safety and final reporting duties continue.

What the pilot does

Legion Health runs an AI supported psychiatry practice. The pilot lets its system authorize a refill of a prescription a Utah patient is already taking, for stable patients on non controlled medication such as common anxiety and depression treatments. The system does not diagnose, does not start a new medication, and does not change a dose.

Relief granted

The Division of Professional Licensing agreed not to bring an enforcement action for unlicensed or unprofessional conduct, or under the state telehealth requirements, where the AI authorizes a renewal that a licensed prescriber would otherwise have had to authorize personally. The relief runs to the specific sections named in Schedule A of the agreement and to nothing else.

What the company agreed to

• A licensed clinician reviews the first 250 requests before any of them completes.
• The next 1,000 requests go through intensive retrospective review, then monthly random sampling and audit continues for the rest of the term.
• Any sign of suicidality, severe side effects, mania, or pregnancy stops the process and sends the case to a clinician. A patient can ask for a human at any point.
• Patients are told AI is part of the process, and identity and prescription are verified before anything is issued.
• Monthly reporting to this office, including agreement rates between the AI and the reviewing clinician.
• Malpractice insurance that covers AI related liability.

What the pilot does

Doctronic uses AI to process 30, 60, and 90 day renewals of medication a licensed provider has already prescribed. It does not write a new prescription, touch controlled or addictive substances, or change a treatment plan. The pilot covers a fixed list of medications set out in the agreement and no others.

Relief granted

The same narrow relief as the Legion Health pilot: the Division of Professional Licensing agreed not to enforce the unlicensed conduct and telehealth provisions named in Schedule A against a renewal the AI authorizes under the agreed protocol. The relief covers the renewal function only and expressly reaches nothing else Doctronic offers.

Where it stands

The pilot is still in Phase 1, which means a licensed medical practitioner authorizes every request. Moving to Phase 2 requires this office to approve it, group of medications by group of medications, and it has not been approved. The 2026 amendment tightened that threshold and removed two medications from the list.

What the company agreed to

• A three phase review, with this office deciding whether each phase advances.
• Identity and prescription verification before any renewal is processed.
• Telling the user, before the conversation starts, that they are interacting with generative AI, how their data will be used, and that they can file a complaint with this office.
• Monthly reporting, including how often the AI and the reviewing physician agreed.
• No selling user data and no advertising against it, deidentified or otherwise.
• Third party adversarial testing of the system.

What the pilot does

Dentacor brings dental hygiene care directly to shelters, recovery programs, and transitional housing, where a supervising dentist is rarely on site. The pilot lets a licensed dental hygienist use an AI radiograph tool to identify three specific conditions, and to perform two specific procedures that follow from them.

Relief granted

Utah normally requires a dentist to supervise a hygienist making those calls. Under the agreement, a hygienist finding and the AI tool reaching the same conclusion stands in place of that supervision, for periodontal disease, complete edentulism, and complete anodontia only, and for scaling and root planing and the fitting of full dentures only. Every other requirement on a dental hygienist still applies.

What the company agreed to

• The hygienist and the AI must reach the same conclusion. If they disagree, the case goes to a licensed dentist.
• Patients consent in writing, and the consent says plainly that no dentist is supervising the procedure and what a hygienist is and is not trained to do.
• No claims to patients about the AI tool beyond what it actually does.
• A testing plan this office has accepted before any patient is seen.
• Monthly reporting covering patients seen, results, incidents, and complaints.
• HIPAA compliant systems, no sale of patient data, and no advertising against it.

What the pilot does

ElizaChat is a Utah company whose app holds guided conversations with students, encourages coping strategies, and is built to recognize when a person needs a human instead. It is not therapy and it does not replace a clinician. The pilot is a phased rollout in school districts that agree to take part.

Relief granted

The concern is that a conversation like this could cross into the practice of mental health therapy, which in Utah requires a license. The agreement does not permit that. What it grants is a 30 day period to fix the problem: if the app does cross the line, ElizaChat must report it, tell the affected users, and remedy it, and the Division holds off on enforcement during that window. The protection stops there. It does not cover a claim by a student, a parent, an injured third party, or any other part of the state.

What the company agreed to

• Written parental consent before any regular use by a student under 18.
• Risk assessment built into the conversation, with high risk users passed to ElizaChat's clinical team and emergency contacts and connected to a licensed therapist in real time.
• Clear disclosure to users that they are talking to AI and what it cannot do.
• Any incident harming a user's health, safety, or finances reported to this office within 24 hours.
• Anonymized storage on HIPAA compliant terms, with the district told how data is used, encrypted, and deleted.
• Whatever additional limits a participating district sets.