Appendix C — What the field expects: AI competency frameworks
A competency is something a doctor should be able to know or do by a given stage of training. A competency framework is a list of them, usually grouped into areas called domains. Medical schools use frameworks to decide what to teach, and accrediting bodies use them to decide what to require.
For AI, there is no single agreed framework yet. Several groups have proposed one, and they agree more than they differ. This appendix lists the ones worth knowing, says where each stands, and shows how this course is built on them.
Use it in three ways:
- When you plan your four weeks, these are the sources your research with ChatGPT should find. Compare its list with this one.
- When a residency program or a school asks what you know about AI, these are the areas they are likely to mean.
- When the AAMC publishes its competencies (below), compare them with your plan and with the coverage table in the preface.
Last checked: October 12, 2026. This area changes fast. Check the links for newer versions.
C.1 The AAMC
The Association of American Medical Colleges (AAMC) represents US and Canadian medical schools. It has three pieces of work on AI.
AI Competencies Across the Learning Continuum. A national set of AI competencies for medical school, residency, and practice, supported in part by the Josiah Macy Jr. Foundation. It is not yet published. The AAMC circulated a first draft for comment in April 2026 and a second in June 2026. Its stated aim is that learners “develop a shared foundation in core AI concepts; understand the ethical, legal, and professional implications of AI-enabled care; and are prepared to effectively use, evaluate, and collaborate around AI technologies”. The same page reports that the share of US and Canadian MD and DO schools teaching AI rose from 53% in 2023 to 77% in 2024. AAMC project page.
Principles for the Responsible Use of AI in and for Medical Education. Version 2.0, July 2025. Seven principles for how schools use AI, such as keeping a human-centered focus, providing equal access, and protecting data privacy. They are rules for institutions, not a list of what you should be able to do. AAMC principles.
AI Competencies for Medical Educators. A framework for faculty, written by an AAMC regional group (the Central Group on Educational Affairs) in 2025. It has eight domains, from understanding AI to continuous professional development. That last domain asks faculty to find their own knowledge gaps and make a learning plan, which is what the planning chapter asks of you. Framework (PDF).
The AAMC’s international committee on AI (IACAI), run through its MedBiquitous program, has also published integration frameworks for medical education, including a matrix of recommendations focused on medical students. IACAI page.
C.2 Other US bodies
Accreditation (LCME). The Liaison Committee on Medical Education accredits US MD programs. In October 2025 it presented a proposed revision of its curriculum standard. The proposed element on patient care would require “education and experiential learning … in the appropriate use of artificial intelligence and other emerging technologies in diagnosis and patient management”. The LCME lists this as a topic not in the 2025 to 2026 standards. It is a proposal, not yet a requirement. Note the word experiential: the proposal asks for practice, not only lectures. LCME presentation (PDF).
The AMA. In November 2025 the American Medical Association adopted policy to develop model AI learning objectives and teaching materials for medical schools. As of this writing it has not published the objectives. Its free seven-part AI in Health Care series on the AMA Ed Hub is written for students. AMA announcement; AMA ChangeMedEd AI series.
The Macy Foundation conference. A 2025 conference with the AAMC and the ACGME, which accredits residencies, made five recommendations. The first is to put AI curricula and competencies in place (“Josiah Macy Jr. Foundation Conference on Artificial Intelligence in Medical Education: Proceedings and Recommendations.” 2025).
C.3 Published frameworks
These were developed by researchers and published in peer-reviewed journals. Each was built by asking experts, by reviewing earlier proposals, or both.
| Framework | How it was made | What it contains | Why it is useful to you |
|---|---|---|---|
| Russell et al., Academic Medicine, 2023 (Russell et al., 2022) | Interviews with 15 experts | 6 domains, 25 subcompetencies, for all health professionals | The most-cited US framework. Its domains are short and readable; start here |
| Hunt et al., npj Digital Medicine, 2026 (Hunt et al., 2026) | Review of 54 studies from 22 countries | 7 domains, 37 competencies, 170 learning objectives | The widest summary of what has been proposed for medical students. Most of its sources were opinion pieces, not studies |
| Lee et al., Academic Medicine, 2024 (Lee et al., 2023) | Two expert rounds plus a national survey of students and faculty in South Korea | 6 domains, 36 competencies | Separates essential from optional: most experts rated data science and AI research optional for graduates |
| Singla et al., npj Digital Medicine, 2024 (Singla et al., 2024) | Three survey rounds with 18 Canadian experts | 82 of 107 competencies rated essential | Includes communication and collaboration, not only knowledge |
| DECODE, Car et al., JAMA Network Open, 2025 (Car et al., 2025) | Two survey rounds with experts from many countries | 4 domains, 19 competencies; 33 required and 145 optional learning outcomes | Covers digital health generally, with AI inside it. Shows how to split a small required core from a large optional set |
Russell’s six domains give a sense of how these frameworks are written. A health professional should be able to:
- explain what AI is and describe its uses in health care;
- explain how social, economic, and political systems shape AI tools, and how that affects justice, equity, and ethics;
- carry out clinical encounters that use AI to build patient-centered care plans;
- evaluate AI tools for quality, accuracy, safety, fit to the setting, and bias;
- analyze and adapt to changes in teams, roles, and workflows when AI tools arrive;
- keep learning about AI tools through continuing education and practice improvement.
C.4 What they agree on
Across these frameworks, the same areas come up again and again:
- Foundations: what AI and machine learning are, how a chatbot produces an answer, and why it can be confidently wrong.
- Checking: judging an AI answer, and the evidence for an AI tool, before relying on it.
- Fairness: where bias comes from, and who a tool may fail.
- Ethics, law, and professionalism: privacy, consent, who is responsible, how tools are regulated, and keeping your own skills.
- Using it in practice: AI in real clinical work, with the doctor still deciding.
- Talking with patients: explaining when and how AI is part of their care.
Data science and building AI tools appear too. Two of the frameworks above (Lee and DECODE) treat them as optional for a medical graduate, and this course does the same.
C.5 How this course uses them
The course’s own list of ten competencies, numbered AIM-1 to AIM-9 plus AIM-X, was built from the frameworks above. Each chapter’s subtitle names the ones it serves, and the coverage table in the preface shows which chapters teach which. The numbers stay fixed so the course can be compared with the AAMC set when it is published.
| ID | Area | Closest match in the frameworks above |
|---|---|---|
| AIM-1 | Foundations of AI and data | Russell 1; Hunt, theory and foundations; Lee 2 |
| AIM-2 | Checking AI tools and their answers | Russell 4; Hunt, critical appraisal |
| AIM-3 | Bias, equity, and social effects | Russell 2; Hunt, ethics |
| AIM-4 | Ethics, law, regulation, and professionalism | Hunt, ethics, law, and professionalism; Lee 3 |
| AIM-5 | Using AI in clinical work | Russell 3; Hunt, clinical applications; Lee 4 |
| AIM-6 | Talking with patients about AI | Singla, communication; Russell 3 in part |
| AIM-7 | Workflow and implementation (optional) | Russell 5 |
| AIM-8 | Health data science (optional) | Lee 5; DECODE, health data science |
| AIM-9 | Research and scholarship (optional) | Hunt, research and innovation; Lee 6 |
| AIM-X | Professional identity and keeping up | Russell 6 |
Talking with patients about AI is the area this course weights most differently from the frameworks. Of the five in Table C.1, only Singla names communication as an area of its own; the others include it, if at all, inside wider domains. The course gives it a core week, because it is a conversation you will have in residency whether or not you use AI yourself.