Doc Groups Respond to AI RFI
Physician groups urge the Trump Administration to pursue standardization, resolve reimbursement and liability concerns, incorporate clinicians' perspectives into policy development, and tackle bias in the creation of AI-enabled tools and other products.
The Trump Administration recently issued a Request for Information (RFI) seeking input on how to advance the use of artificial intelligence (AI) in clinical care. The Department of Health and Human Services (HHS) asked stakeholders for feedback on how it can utilize regulatory, reimbursement, and research policies to promote AI in clinical settings. Over 7,000 comments were received by the deadline of February 23, 2026.
To evaluate the responses from the physician community, I reviewed comments from the American Academy of Family Physicians (AAFP), the American College of Cardiology (ACC), the American College of Radiology (ACR), the American College of Surgeons (ACS), the American Medical Association (AMA), and the Medical Group Management Association (MGMA). I chose ACC and ACR because these specialties are currently heavily impacted by AI, especially in image analysis. While the responses from all these groups reflected their respective memberships' priorities, common themes recurred throughout the letters. Highlights are discussed below.
Highlights from Letters Submitted by AAFP, ACC, ACR, ACS, AMA, and MGMA
The Trump Administration’s stated approach to AI can be simply described as pro-innovation and deregulatory. It favors a top-down strategy in which the federal government sets the rules of engagement. They have attempted to disrupt legislative efforts in individual states and have shown little or no interest in engaging with certain professional collaborative groups, such as the Coalition for Health AI (CHAI), which worked closely with the Biden Administration.
While the comment letters I reviewed mostly aligned with this approach, they also diverged in some important ways. These differing approaches were based on advancing AI according to scientific principles; a broader vision for the medical profession and patient care; and a practical, real-world understanding of the clinical environment. Five key points from my review of these letters are outlined below.
#1: Variety of AI: Commenters recommended that regulatory approaches and policies should account for differences across AI categories. Medical practices use AI in several ways, and specialty societies urged the Administration to develop policies that reflect the range of applications and their related risks. AI can be used for administrative tasks (e.g., documentation, scheduling, billing) as well as for direct patient care. In direct medical care, it can support automated functions (e.g., robotics) or provide more detailed support, such as data analysis to inform diagnosis and treatment decisions (e.g., software-based systems).
#2: Role of the Physician in AI Policymaking: There was unanimous agreement that physicians should be involved in all aspects of AI development, deployment, and ongoing oversight. Physicians are viewed as best positioned to inform both the clinical aspects of AI and assess how AI is used in medical practice. The AMA, in particular, emphasized that for prior authorization decisions, AI can support the process, but “adverse coverage decisions must not be fully automated. Denials should require timely review by a qualified, licensed physician, with escalation pathways to protect patient safety when care may be delayed.” The importance of clinical decision-making was also emphasized, and it was recommended that it should remain in the hands of the clinician and the patient to determine the best course of action. Commenters urged HHS to collaborate with medical specialty societies as HHS develops and implements AI-related policies.
#3: Critical Elements of the Overall Regulatory Environment: Several common themes emerged, recommending a unified government approach that prioritizes standardization, transparency, and the minimization of bias in data inputs.
Unified approach: There was general agreement that a unified strategy across the federal government was necessary, and some concerns were raised about the risk of too many conflicting state-level regulations. While this generally aligns with the Trump Administration’s preferred top-down approach, the societies also urged the federal government to collaborate meaningfully with a wide range of stakeholders. An approach they have largely avoided.
Standardization: Commenters highlighted the importance of standardization across AI products and policies. They pointed out that semantic standardization is crucial for responsible AI use. It was recommended that AI training, usage, and outputs should be based on standardized concepts and code sets. Such a framework would require cooperation among various groups, such as EHR vendors, payers, and AI product developers. The importance of privacy and cybersecurity accountability was also discussed. However, this recommendation might not fully match the Administration’s expressed preference for a deregulatory approach.
Transparency: The importance of enhancing transparency across all areas of AI development and deployment was highlighted. Greater transparency can help inform AI users, such as physicians, and also build greater public trust in AI systems. It was suggested that this effort could be supported by the federal government or accredited organizations. Two respondents expressed disappointment with a recent HHS regulation that limited the information available to users of AI-enabled decision support tools. Commenters recommended that patients, clinicians, and other stakeholders have a clear understanding of, and access to, the datasets used to train AI models.
Addressing Bias: Overall, commenters supported efforts to identify and reduce bias in AI-enabled tools, medical devices, and other applications. Bias in data inputs can produce biased or incorrect results. Reducing bias aligns with scientific principles for research and data collection. In healthcare, this could lessen the value and usefulness of AI outputs and harm public trust. These comments conflict with the Trump Administration's policies. The Trump Administration sought to eliminate Diversity, Equity, and Inclusion (DEI) initiatives, calling them "divisive concepts" and "un-American."
#4: Reimbursement and Liability Policies: Commenters pointed out that the absence of a clear reimbursement pathway and uncertainty about liability risks were barriers to AI adoption. They recognized that the changing nature of AI makes it hard to establish these policies, creating a barrier to progress.
Reimbursement: Commenters focused on finding solutions within a fee-for-service framework and developing approaches within a value-based model. Regardless of the approach chosen, commenters expressed concerns about the impact of high AI costs in a budget-neutral environment, such as the Medicare Physician Fee Schedule (PFS). Covering the rising costs of AI could lead to reductions in payments for other services.
Liability: Commenters observed that current liability laws might not apply to AI, and liability issues related to AI in medicine remain unresolved. The diversity of AI applications means liability risks will vary by specific AI product. There was concern that responsibility does not fall entirely on physicians and medical practices, and that the liability of AI developers should also be considered. Commenters urged HHS to take a leadership role in developing guiding principles that promote accountability and lessen uncertainty. Most commenters agreed on concepts such as dividing liability between users and developers, HHS taking the lead in establishing liability policies, and the liability framework evolving as AI technology advances.
#5: Specialty Societies' Engagement in AI: In their letters, specialty societies highlighted their efforts to support members with emerging AI issues. For example, the AMA outlined the CPT Editorial Panel’s work on updating CPT codes by creating a unique coding framework that precisely describes algorithmic services within the patient care pathway. The ACR established the Data Science Institute in 2017 to promote safe, effective, and clinically useful AI innovations for radiologists. The AAFP is creating the Primary Care Innovation Network to develop a sustainable, physician-led mechanism that links frontline primary care teams with AI and digital health solution developers. By promoting these initiatives, specialty societies are demonstrating to HHS how they can be effective partners in addressing critical issues in advancing AI in the clinical environment.
Now that HHS has received responses to its RFI, we might see it offer more insights into its approach to advancing AI. Congress has not yet enacted any significant AI legislation. Currently, they are focused on addressing the significant energy demands of AI data centers—expected to double or triple by 2028 and potentially make up 12% of U.S. electricity use. Still, this Administration has shown a willingness to push policies quickly, so we could see announcements from HHS soon. Some policies are likely to be included in the 2027 Medicare PFS Proposed Rule released around July 1, but action could come even sooner. The question remains whether HHS will consider guidance from this RFI that may not fully align with the Administration’s stated goals, such as deregulation.
__________________
For more information and questions, please contact:
Sheila Madhani
Madhani Healthcare Consulting
Email: smadhani@madhani-health.com
Tel: (202) 679-2977