American Medical Association

The American Medical Association (AMA) is the United States' oldest and largest physician organization, representing approximately 271,000 physicians and medical students through a federated governance structure of more than 190 state and specialty medical societies. Operating from Chicago with a Washington D.C. policy presence, the AMA's pivot into AI governance follows the same sectoral-pivot pattern the corpus tracks elsewhere: a long-established professional association whose foundational commitments — physician authority in clinical decision-making, patient safety, health equity — are directly threatened by AI deployment in ways that reorganize the association's advocacy agenda. The AMA's AI intervention operates on two axes: a clinical-integrity axis, establishing that AI must augment rather than replace physician judgment and must be validated, transparent, and free of discriminatory design; and a market-power axis, contesting insurance carriers' use of automated systems to deny care with minimal human review, which the AMA frames as both a patient-safety threat and an assault on physician authority.

Founding and structure

The AMA was founded on May 7, 1847, in Philadelphia by Nathan Smith Davis and a convening of more than 250 physicians from forty medical societies and twenty-eight medical colleges — the first national effort to establish scientific and ethical standards for medical practice in the United States. Its founding mission centred on raising standards in medical education and practice. By 2022 membership stood at approximately 271,000, making it the largest professional physician organization in the country.

The AMA's governance is bicameral. The House of Delegates — modelled after the U.S. House of Representatives — convenes twice a year with elected representatives from more than 190 state and specialty medical societies; it functions as the AMA's legislative body, adopting policy positions that shape both internal governance and external advocacy. The Board of Trustees provides executive oversight and approves advocacy positions between sessions. Operationally the AMA maintains a Chicago headquarters at 330 North Wabash and a Washington D.C. policy office; John Whyte, MD, MPH serves as CEO and Executive Vice President.

The AMA's advocacy reach operates at multiple scales simultaneously: its Washington office channels House of Delegates policy into Congressional testimony, federal rulemaking comments, and direct engagement with executive agencies; its relationships with 190+ specialty societies amplify its positions through every clinical specialty in U.S. medicine.

Terminology: augmented intelligence, not artificial intelligence

The AMA uses "augmented intelligence" rather than "artificial intelligence" as its standard term for AI in healthcare — a deliberate framing that emphasises the assistive and supportive role the technology is expected to play. In the AMA's formulation, AI is designed to "enhance human intelligence rather than replace it," and the measure of a well-designed health AI system is whether it supports physician judgment, not whether it can substitute for it. This terminological commitment is not merely rhetorical — it grounds the AMA's entire governance framework: validation requirements, transparency obligations, liability standards, and resistance to autonomous AI decision-making in clinical contexts all follow from the premise that AI is an assistant and the physician remains the responsible agent.

The Council on Long Range Planning & Development (CLRPD) formalised this framing in two internal research documents: A Primer on Artificial and Augmented Intelligence, which provided the AMA membership with historical context, definitions, and an assessment of where AI stood in healthcare and where it might go; and Generative AI in Medicine and Health Care, a follow-on primer addressing the distinctive properties of large language models — architecture, terminology, prominent deployed models, clinical promises, and failure modes. Both documents exist as resources for the AMA's policy-development process, establishing the epistemic foundation for the formal policies the House has since adopted.

Core AI policies — H-480.940 and H-480.939

The AMA's entry into AI governance policy dates to its June 2018 Annual Meeting, when the House of Delegates adopted Policy H-480.940, "Augmented Intelligence in Health Care." This was the AMA's first comprehensive policy framework for AI in medicine. H-480.940 establishes that health care AI should be: thoughtfully designed and high-quality; clinically validated through best practices in user-centred design; transparent in its operation; reproducible per established standards; proactively de-biased, with attention to disparate impacts on vulnerable populations; and protective of patient privacy and data security. The policy explicitly enumerates values of ethical relevance — professionalism, transparency, justice, safety, and privacy — as co-equal concerns alongside technical performance standards.

In 2019, the House adopted Policy H-480.939, also titled "Augmented Intelligence in Health Care," which addresses the deployment side: the AMA supports use and payment of AI systems that advance the quadruple aim — improving the patient experience of care, improving population health, reducing healthcare costs while increasing value, and improving the professional satisfaction of physicians and the care team. The quadruple aim framing is significant: it grounds AI deployment legitimacy in both patient outcomes and physician experience, making degradation of the care team's professional environment — including AI-driven workload shifts or deskilling — a policy concern, not merely a labour dispute.

Together the two policies established a governance position that preceded the generative AI era: the AMA was building an ethical framework for AI in healthcare before large language models entered clinical settings, which gave it a principled baseline to apply when they did.

November 2023 AI advocacy principles

At its November 2023 Board of Trustees meeting, the AMA adopted a set of seven advocacy principles for AI development, deployment, and use — intended as a framework for engagement with Congress, federal agencies, and the technology industry:

1. Oversight — A whole-of-government approach to AI governance, while acknowledging the roles of non-governmental entities.
2. Transparency — Mandated disclosure of AI system design, development, and deployment characteristics.
3. Disclosure and documentation — Required when AI materially affects patient care or medical decisions.
4. Generative AI — Governance policies must be in place before adoption and use, not after.
5. Privacy and security — A privacy-by-design approach with robust safeguards on health data.
6. Bias mitigation — Proactive identification and elimination of algorithmic bias in health AI systems.
7. Liability — Explicit protections for physicians' liability exposure when using AI-enabled technology.

The liability principle reflects a distinctive structural concern: where AI systems are required or heavily incentivised by health systems or insurers — particularly for prior authorization or clinical-decision support — physicians are increasingly exposed to adverse outcomes produced by systems they did not design, cannot fully audit, and in some cases cannot override. The AMA's position is that liability frameworks must account for this structural reality rather than simply imputing responsibility to the clinician at the point of care.

AI in prior authorization — the market-power axis

The AMA's most sustained public campaign on AI concerns its use by health insurers to deny or delay care through automated prior authorization systems. Prior authorization — the requirement that insurers approve certain treatments, procedures, or medications before a physician can proceed — has long been a point of friction between organized medicine and the insurance industry. The emergence of AI-driven automation in denial decisions elevated the conflict to a new register.

In June 2023, concurrent with the House of Delegates' directive to develop AI principles, the AMA documented a specific case: Cigna's use of an automated review system in which more than 300,000 claims were denied with employed physicians spending an average of 1.2 seconds per claim — effectively removing meaningful human review from the process entirely. The AMA House of Delegates directed the organisation to advocate for federal oversight ensuring that insurer AI systems: base decisions on accurate clinical criteria drawn from medical specialty guidelines and peer-reviewed literature; include review by physicians without financial incentives to deny care, possessing relevant clinical expertise; require actual human examination of patient records before any denial; and are subject to transparent evaluation for potential biases. AMA data from this period indicated that 1 in 3 physicians had witnessed prior authorization causing serious adverse events for patients.

The AMA's position is that medical necessity determinations are clinical acts, not administrative ones, and that algorithmic substitution for physician judgment in denial decisions is both a patient-safety problem and a structural violation of the physician-patient relationship. This axis connects the AMA's AI governance work to its longstanding prior authorization reform campaign and gives the AI-governance agenda real-stakes grounding in practitioner experience.

Physician sentiment surveys and the Center for Digital Health and AI

The AMA has tracked physician attitudes toward AI through repeated sentiment surveys since at least 2023, providing an empirical baseline for its policy advocacy. The survey data show rapid adoption alongside persistent anxiety: in 2023, 38% of physicians reported using AI tools in their practice; by 2024 the figure had risen to 66%; by 2026 it exceeded 80%. The top opportunity physicians cite for AI is reduction of administrative burden — documentation, billing, prior authorization — accounting for a substantial share of the physician dissatisfaction the quadruple aim is meant to address. The primary concerns centre on data privacy, reliability, and the risk of AI-generated clinical recommendations that are confidently stated but clinically incorrect.

On October 20, 2025, the AMA launched the Center for Digital Health and AI, a new organisational unit announced by CEO John Whyte, MD, MPH, after two years of studying how physicians interact with AI systems. The Center's mandate has four focus areas: policy and regulatory leadership, shaping standards for safe and effective AI in medicine through engagement with regulators, policymakers, and technology industry leaders; clinical workflow integration, creating structured pathways for physicians to shape AI tools so they fit within clinical practice; education and training, providing physicians and health systems with practical guidance on AI implementation; and cross-sector collaboration, building partnerships across healthcare, technology, research, and government. The Center carries forward the AMA's AI Specialty Collaborative — bringing together 21 medical specialty societies — as a standing mechanism ensuring that specialty-specific clinical perspectives, not only generalist positions, shape health AI standards and governance.

Position in the corpus

The AMA's place in the make-AI-good corpus is as the organised voice of the US physician community on AI governance — the professional association whose foundational commitments most directly collide with the ways AI is being deployed in clinical and administrative medicine. Its two axes — clinical integrity (AI must augment, not replace, physician judgment) and market power (insurers must not use AI to substitute automated processing for physician-reviewed decisions) — together produce a governance agenda that is simultaneously pro-responsible-AI and explicitly adversarial toward AI deployments that degrade care quality, deskill physicians, or create liability exposure without corresponding accountability.

Structurally, the AMA follows the sectoral-pivot pattern the corpus tracks across professional communities — the American Library Association, the National Education Association, the Authors Guild — where a long-established professional association pivots into AI advocacy when AI deployment directly reshapes its members' professional domain and threatens the institutional commitments the association exists to protect. The AMA's distinctive feature in this pattern is scale and institutional reach: with 271,000 members, a House of Delegates spanning 190+ specialty societies, and a Washington office with standing access to federal rulemaking processes, the AMA's pivoted AI agenda connects directly to the legislative and regulatory levers of American healthcare.