Regulatory measures and policy guidance on AI applications in health care have emerged across state governments, professional medical bodies, and academic publishers, emphasizing the need for human supervision and responsibility. Concurrently, recent surveys indicate that a growing number of adults would accept AI-driven health tools provided that a qualified professional reviews the final output. Separate federal actions aim to broaden studies of potential vaccine-related health issues, though concerns persist about how these efforts may be perceived amid skepticism toward government health agencies.
A Transparency Coalition analysis identified 115 AI-related legislative actions in 30 states this year, with 108 enacted and seven pending gubernatorial approval as of mid-October. The report highlighted human oversight in medical AI and mandatory disclosures as the most frequently addressed topics. Among the new laws, eight states now restrict insurers from using AI alone to reject or reduce coverage, including automated prior authorization decisions.
At least five states have enacted laws this year restricting AI in mental health care, generally requiring therapy to be provided by a licensed human professional. California’s legislation further requires health facilities to ensure clinicians maintain independent judgment when deploying AI-assisted diagnostic tools. Most mental health AI restrictions are already active, while Rhode Island’s and California’s provisions begin January 2027.
State Laws and Professional Guidance
Legislative activity at the state level has paralleled voluntary initiatives from medical boards and professional associations, all prioritizing human accountability in AI-driven health care. In October, the American Medical Association (AMA) published a governance framework for autonomous AI systems, outlining key questions for clinical applications—such as the evidence supporting their use and the parties responsible for associated risks. The framework serves as a foundation for the AMA’s future work on physician liability, licensing requirements, and regulatory standards, though it does not impose mandatory rules on doctors or AI developers.
Meanwhile, the Federation of State Medical Boards (FSMB) has formed a working group to draft recommendations for state medical boards regarding the supervision of AI tools performing clinical functions with minimal physician involvement. FSMB leaders have consistently argued that existing medical accountability frameworks should govern AI use, with physicians retaining ultimate responsibility for AI-assisted decisions. Draft recommendations are expected shortly and will undergo public review before potential adoption, though they carry no legal enforceability unless incorporated into state policies.
Academic Publishers Set New Rules
In October, Health Affairs—a leading health policy journal—adopted an AI usage policy aligning with standards adopted by many academic publishers since 2023. The policy explicitly prohibits authors from crediting AI as a co-author or substituting it for human judgment in scholarly work. Authors must personally verify AI-assisted content, take full responsibility for its accuracy, and disclose any AI involvement. Additionally, the journal advises against relying on AI for citation management, citing frequent inaccuracies in generated references.
These publishing guidelines reflect broader concerns about the role of human expertise in AI-assisted health care, research, and policy development. Clinicians, researchers, and regulators alike are grappling with how to integrate AI while preserving accountability. Public opinion surveys further suggest that transparency and human oversight remain critical factors in acceptance of AI in personal health care.
Expanding Vaccine Safety Research
The Department of Health and Human Services (HHS) recently launched initiatives to enhance research and reporting on health issues potentially linked to vaccination. Key components include establishing a dedicated vaccine injury clinic at the National Institutes of Health (NIH) Clinical Center. Federal officials framed the efforts as a response to patient reports of being dismissed when raising concerns about vaccine-related health problems. A 2024 New York Times investigation, for example, detailed accounts of individuals experiencing symptoms after COVID-19 vaccination who felt ignored by healthcare providers and government agencies. Research in the Veterans Affairs system has similarly shown that dismissive attitudes toward vaccine safety concerns can damage trust between patients and clinicians over time.
However, the NIH Clinical Center’s governing board has expressed reservations about how the new clinic may be perceived, warning that findings could be met with skepticism regardless of their validity. The board’s statement noted that the clinic’s opening ceremony, featuring discussions with individuals claiming vaccine injuries, might imply premature conclusions. While acknowledging that rare but serious vaccine injuries warrant rigorous study, the board emphasized the importance of clear communication about research methods and limitations.
Some medical groups, including the Infectious Diseases Society of America (IDSA), have criticized the proposed Medicare billing code, arguing it is unnecessary and could distort public perception by suggesting vaccine injuries are more common than they are. Existing federal systems, such as the Vaccine Adverse Event Reporting System (VAERS), a joint FDA/CDC passive surveillance program, already track and investigate potential vaccine-related health issues, though VAERS reports are unverified and do not prove causation.
Establishing Standards for AI Use
The AMA’s governance framework for autonomous AI in clinical settings poses foundational questions about evidence, oversight, and accountability-without imposing mandatory requirements on physicians or developers. Similarly, Health Affairs‘ policy reinforces the principle that AI must supplement, not replace, human judgment in academic research. Both efforts show the need for rigorous validation of AI-generated content, particularly in high-stakes fields like medicine and public health.
