By admin 
The safest and most sensible answer today is that AI is not ready to be independently licensed like a physician. To understand why, one must look beyond the technical capabilities of Large Language Models (LLMs) and examine the bedrock upon which modern medicine is built. Medicine rests on a profound social contract. For centuries, society has granted physicians unique, near-monopolistic privileges—such as the authority to prescribe controlled substances, perform invasive surgeries, and make life-altering diagnoses—because they promise extraordinary duties of competence, ethics, and accountability in return. A medical license, therefore, isn’t just a permit to generate thoughtful and informed answers or to pass a standardized exam like the USMLE. It is a legal and ethical grant of authority tied to human judgment, professional accountability, rigorous discipline, radical transparency, and an unwavering fiduciary duty to put patients first.
The Federation of State Medical Boards (FSMB), which represents the 69 state and territorial medical boards in the United States, occupies a unique position in this unfolding drama. While its role is primarily advisory, the FSMB provides the essential framework that allows state boards to fulfill their statutory charge: licensing and regulating physicians to protect the public. From this vantage point, the FSMB views generative artificial intelligence as a powerful tool that should be governed within existing professional and institutional accountability structures. The prevailing philosophy is that responsibility must be aligned with the level of autonomy and human oversight, rather than treating the software as a separate “practitioner” licensed in the same manner as an individual clinician.
The challenge for regulators is to help boards navigate this emerging landscape by ensuring that regulation keeps pace with innovation while promoting consistent, ethical standards. At the same time, no responsible regulator can or should ignore the nation’s health care workforce crisis. The United States is facing a projected shortage of up to 86,000 physicians by 2036, according to the Association of American Medical Colleges (AAMC). In this context, well-governed AI tools could help relieve staggering administrative burdens, improve rural access to care, and allow clinicians across many specialties to spend more meaningful time with their patients by automating the "drudge work" of documentation and data entry.
In 2024, the FSMB’s House of Delegates took a significant step by adopting a nonbinding policy guidance regarding the responsible incorporation of AI into clinical practice. The message—intended for medical regulators and licensed clinicians alike—was designed to be lofty and overarching, but balanced and practical. It encourages the use of AI by physicians when it can demonstrably improve care and advises clear documentation of its use. However, it also issues a stern caution: professional responsibility remains entirely with the human licensee. State medical boards do not regulate machines or tools; they lack the statutory authority and technical infrastructure to do so. They regulate the human beings who use those tools. Consequently, state medical boards maintain that physicians should remain fully accountable for any harms caused by an inappropriate reliance on AI.
This principle of human-centric accountability is under pressure as some policymakers begin to imagine a separate lane for licensing autonomous systems. In 2024, proposed legislation in Idaho and Iowa raised the possibility of creating a state licensing board for “autonomous service providers.” These proposed entities would have sat apart from each state’s long-standing medical board, effectively creating a regulatory "silo" for non-human practitioners. While both bills ultimately failed to pass, their introduction reveals a tectonic shift in the political landscape. We are past the point of wondering whether someone will propose licensing AI like a human physician; the proposal has already been made. The question now is who will decide the future of these systems, under what authority they will operate, and whether the decision-makers have a legitimate understanding of the practice of medicine and the cascading impacts that autonomous AI brings to patient safety.
There is a profound danger in treating medicine as just another commodity—a consumer service that can be split off into a new regulatory silo for the sake of efficiency or market competition. Medicine is fundamentally different because the stakes are existential. A prescription refill, a triage recommendation, or a diagnostic suggestion may look routine or "low-risk" until it is handled poorly, leading to a catastrophic drug interaction or a missed oncology diagnosis. The public deserves clear lines of responsibility and a guaranteed path for recourse when harm occurs. If a human doctor errs, there is a clear process for investigation, license suspension, and legal liability. If an "autonomous service provider" errs, the lines of accountability become blurred in a thicket of software end-user license agreements and corporate indemnity.
Recent events in Utah illustrate both the immense promise of AI and the friction that occurs when technology moves faster than regulation. The state’s Office of Artificial Intelligence Policy recently approved a 12-month pilot program with a health tech startup called Doctronic. The goal was to automate routine, guideline-based prescription renewals for 192 commonly prescribed drugs under strict parameters and physician oversight. While the program was framed as a way to increase efficiency, it quickly drew the attention and concern of Utah’s medical licensing board. This friction underscores a critical point: even when innovation moves through a state’s dedicated AI apparatus or "regulatory sandbox," the medical board must have a primary seat at the table whenever clinical decisions and patient safety are involved.
The FSMB’s 2024 policy explicitly advised states exploring the application of AI in health care to work in close partnership with their medical boards. The policy also suggested that boards should proactively examine how the “practice of medicine” is defined in their specific jurisdictions. As AI systems become more "agentic"—meaning they can take independent actions rather than just providing suggestions—the legal definition of practicing medicine becomes the front line of the debate. If an AI system independently modifies a dosage based on real-time biometric data, is it "practicing medicine"? If so, how can that practice be regulated if the "practitioner" has no physical presence, no conscience, and no ability to be disciplined by a board?
This partnership between innovation and oversight is evolving at a breakneck speed. The FSMB is currently working with its member boards, stakeholder organizations, bioethicists, and AI experts to update its guidance. This update is necessary because the existing framework was drafted before the newest wave of agentic systems pushed AI from passive assistance toward more autonomous action. In the early days of medical AI, tools were largely limited to "Computer-Aided Detection" (CAD) in radiology, where a machine would circle a suspicious area for a human to review. Today, we are seeing systems that can draft entire clinical notes, engage in empathetic-sounding conversations with patients about their symptoms, and suggest complex treatment plans.
The debate over AI licensing will not stay confined to academic conference panels or medical boardrooms. Federal and state lawmakers, governors, technology offices, and the general public are already asking who should regulate these increasingly autonomous systems. The answer must be informed by the institutions that already possess a deep, historical understanding of professional standards, clinical risk, and the nuances of the patient-physician relationship. From the FSMB’s perspective, this means carefully distinguishing between the act of "licensing" a human clinician and the act of "authorizing," "registering," or "approving" an AI tool.
As the Washington State AI Task Force and other similar bodies begin their work, they are facing the complex questions raised when AI begins to perform clinical tasks with limited or no direct physician supervision. These are not just technical hurdles; they are sociological ones. As the sociologist William Bruce Cameron wrote in his 1963 book, "Informal Sociology: A Casual Introduction to Sociological Thinking," “Not everything that counts can be counted, and not everything that can be counted counts.” He wrote those words during a corporate mainframe computing boom, long before the era of neural networks. His insight remains relevant today: while we can measure the accuracy of an AI’s diagnostic output, we cannot easily "count" the value of the trust, empathy, and moral responsibility that a human physician brings to the bedside.
As AI systems grow more capable and their outputs become more indistinguishable from those of humans, the central task of medical regulators will be to preserve clear lines of human accountability. The medical community must ensure that innovation serves the patient, rather than the patient becoming a data point for the innovation. By maintaining the medical license as a unique, human-centric bond, regulators can ensure that grounded judgment and patient trust remains the heart of the healing arts, even in an era of unprecedented technological change. The goal is not to stifle the evolution of medicine, but to ensure that as we move into the future, we do not leave behind the ethical foundations that make medicine a profession rather than just an industry.
