State Medical Boards Should Oppose AI-Enabled Prescription Renewal Systems
Humayun Chaudhry, DO, MACP
President and CEO
Federation of State Medical Boards
499 Fuller Wiser Road
Euless, TX 76039
Dear Dr. Chaudhry,
The undersigned 11 organizations representing diverse interests and sectors urge the Federation of State Medical Boards to push for state medical boards to oppose artificial intelligence (AI)-enabled prescription renewal systems pending further review of their legality, safety, accountability, and compliance with existing standards governing the practice of medicine.
This issue arose most recently in Utah, where the state’s Office of Artificial Intelligence Policy is moving forward with a pilot for such a system even though the Utah Medical Licensing Board has called for suspending the pilot.[1] State medical boards should exercise heightened vigilance regarding the expanding use of artificial intelligence systems in clinical decision-making, particularly when AI tools are being positioned to influence prescription renewals, diagnoses, treatment recommendations, triage, or other functions traditionally reserved for licensed medical professionals.
Recent state experimentation with AI-enabled prescription renewal systems has accelerated a broader policy question that now confronts every medical board and AI oversight authority in the country: whether AI marketed as efficient clinical support is now moving beyond assistance and into the unauthorized practice of medicine.
Some pilot efforts have been described as narrow or carefully supervised, but even limited deployments reveal how quickly temporary oversight mechanisms can evolve into autonomous decision-making with diminished physician involvement. As recent medical scholarship has noted, several pilot designs begin with only physician review for an initial subset of cases before transitioning toward independent AI action, creating a pathway in which human review becomes episodic rather than continuous.[2]
Prescription renewal is not an administrative exercise that can safely be treated as low risk for automation. Prescription renewals often require individualized judgment regarding whether a patient’s clinical condition has changed, laboratory work is overdue, adverse side effects have emerged, interactions with newly added medications exist, or continued prescribing remains medically appropriate under evolving circumstances.
Scholars reviewing Utah’s prescription-renewal pilot have emphasized that meaningful therapeutic monitoring, dose adjustment, and clinical reassessment remain core medical functions better suited to licensed professionals than to software systems operating under generalized assumptions.[3]
There is legitimate concern that continued deployment of these systems, while unresolved legal and clinical concerns remain, could normalize AI systems performing core medical functions before adequate safeguards, accountability structures, and independent evidence are in place. Suspending current efforts and halting future ones is therefore an appropriate and necessary precautionary measure while regulators, medical boards, and policymakers determine whether such systems can lawfully and safely operate within existing medical practice frameworks.
Experience across sectors demonstrates that when AI generates a recommendation, professionals often face institutional pressure to approve outputs quickly, particularly in high-volume settings.[4] In healthcare environments, this creates a serious risk that nominal physician supervision becomes little more than rubber-stamp approval of machine-generated conclusions. Systems advertised as “physician supervised” may, in practice, normalize deference to algorithmic outputs where speed, workflow efficiency, and labor reduction are prioritized by administrators. Research has further demonstrated that excessive reliance on AI systems can contribute to skill degradation and automation bias among medical professionals.[5]
There is also little independent evidence supporting the safety and efficacy of these systems under real-world prescribing conditions. In several recent cases, publicly cited validation has relied on simulation studies, company-authored preprints or internal comparisons conducted in settings materially different from those in which these tools are now being deployed. Independent clinical evidence demonstrating safety in chronic medication-renewal contexts remains limited.[6]
Medical boards and AI regulators should also consider the downstream consequences for accountability and patient protection. When an AI-generated clinical decision causes harm, patients may struggle to determine who bears responsibility. Emerging scholarship has raised concerns that some AI vendors reserve broad contractual protections for themselves while patients face substantial barriers to obtaining records needed to understand what occurred or who may be legally accountable.[7] Without reliable audit trails, preserved decision logs, and clear documentation of how outputs were generated, disciplinary review and patient protection become substantially impaired.
Equally concerning is the growing use of marketing language that equates AI software with licensed medical practice. Terms such as “AI doctor,” “medical-grade AI,” or similar descriptors risk misleading patients into believing they are receiving care from an entity possessing legal licensure, professional judgment, and fiduciary obligations equivalent to those of a physician.
No software application holds a medical license. No AI system assumes professional ethical duties. No algorithmic model can independently satisfy the obligations imposed on licensed medical professionals. This creates a dangerous discrepancy in which licensed clinicians remain subject to professional liability while AI health companies seek to avoid equivalent accountability.
We are calling for the Federation of State Medical Boards to push state medical boards to adopt the following principles when evaluating AI deployment in clinical settings:
- AI must not independently perform acts that constitute the practice of medicine absent clear physician accountability. Prescription initiation, renewal, modification, diagnosis, and treatment recommendations should remain attributable to a licensed clinician who meaningfully reviews patient-specific circumstances.
- Physician oversight must be substantive, not symbolic. Boards should require documentation demonstrating when clinicians intervene, override, or reject AI outputs rather than assuming oversight exists because a physician is nominally connected to a system.
- Patients must receive clear disclosure when AI is involved in any clinical recommendation. Disclosure should explain that AI is a software tool, identify the extent of physician review, and avoid marketing language implying professional licensure.
- Medical boards should require preserved records sufficient for investigation. Any AI-assisted clinical interaction should generate accessible documentation showing what information the system relied on, what recommendation it produced, and how a clinician responded.
- Boards should coordinate with federal regulators where device oversight may be implicated. Several legal scholars have argued that autonomous prescribing systems likely fall within the category of medical devices requiring federal scrutiny, particularly when software is performing treatment functions rather than merely supporting clinician judgment.[8]
Although medical professional shortages and healthcare system strain are serious concerns, pressure to improve efficiency cannot justify weakening foundational safeguards governing patient safety, long-term disease management, informed medical judgment, and clinical accountability.
State medical boards exist precisely because medical expertise, clinical oversight, and professional accountability require specialized knowledge developed through licensure, training, and ongoing regulatory supervision. State bodies with expertise in technology policy, but not medical licensing authority, should defer to the professional judgment of the relevant state licensing boards on questions involving the practice of medicine and patient safety.
We cannot permit a regulatory framework in which generalized AI expertise is treated as sufficient to supersede the judgment of trained professionals and expert oversight bodies within their respective fields. Allowing technology officials or AI developers to override medical boards on matters of clinical practice risks eroding longstanding safeguards designed to protect patients from unsafe or insufficiently tested interventions. That responsibility is especially important now, before autonomous clinical systems become normalized through premature deployment and experimentation.
We urge the Federation of State Medical Boards to support efforts to ensure that AI systems do not displace the professional obligations, judgment, and accountability that licensed medical practice requires.
Sincerely,
AFT: Education, Healthcare, Public Services
GLMA
Labor Campaign for Single Payer
National Health Care for the Homeless Council
National Health Law Program
National Association of Social Workers
People’s Action Institute
Physicians for a National Health Program
Popular Democracy
Public Citizen
Social Security Works
Sources:
[1]Letter from Alan Smith, Chair, Utah Medical Licensing Board, to Utah Department of Commerce, Office of Artificial Intelligence Policy (April 20, 2026), https://bit.ly/42VVxnO.
[2]Michelle M. Mello, Utah’s Experiment With AI-Driven Prescription Renewals, 7 JAMA Health Forum 1, 1-3 (2026).
Sara Gerke et al., Utah’s Prescription-Renewal Pilot Program — Autonomous AI Managing Patient Care, 394 The New England Journal of Medicine, 1561, 1561-1563 (2026).
[3]Sara Gerke et al., Utah’s Prescription-Renewal Pilot Program — Autonomous AI Managing Patient Care, 394 The New England Journal of Medicine, 1561, 1561-1563 (2026).
[4]Dimitris Giannitsas et al., In Artificial Intelligence (AI) We (Dis)trust? Navigating Institutional Pressures for Automation and Augmentation in the Implementation of AI in Organizations, 36 Information and Organization 1, 1-25 (2026).
[5]Moustafa Abdelwanis et al., Exploring the Risks of Automation Bias in Healthcare Artificial Intelligence Applications: A Bowtie Analysis, 5 Journal of Safety Science and Resilience 460, 460-469 (2024).
[6]Daniel G. Aaron and Christopher Robertson, The First AI Drug Prescriber, JAMA Viewpoint (April 13, 2026), https://bit.ly/4tkaj2o.
[7]Michelle M. Mello, Utah’s Experiment With AI-Driven Prescription Renewals, 7 JAMA Health Forum 1, 1-3 (2026).
[8]Michelle M. Mello, Utah’s Experiment With AI-Driven Prescription Renewals, 7 JAMA Health Forum 1, 1-3 (2026).
Sara Gerke et al., Utah’s Prescription-Renewal Pilot Program — Autonomous AI Managing Patient Care, 394 The New England Journal of Medicine, 1561, 1561-1563 (2026).