AI can support medical learning and clinical work, but it should not lower the bar for what a physician must know and be able to do. Medical training should prepare clinicians to use AI critically while still demonstrating the judgment and skills needed when a system is unavailable, unreliable, biased, or poorly suited to a case. “The physician is the only redundancy” is a call for resilient human competence and accountability—not a claim that AI has no value or that physicians are the only safeguard against harm.
What competency means when AI is available
Competency-based medical education (CBME) defines and assesses outcomes through competencies, or observable abilities, with the aim of achieving patient-centered outcomes. That distinction matters when AI enters a learning or clinical workflow: access to a tool is not itself evidence that a learner can reason through a case, make a decision, communicate it, or recognize when a recommendation is unsafe.
As an Amazon Associate I earn from qualifying purchases.
The Association of American Medical Colleges (AAMC) says its foundational undergraduate medical education competencies, developed with the American Association of Colleges of Osteopathic Medicine and the Accreditation Council for Graduate Medical Education, were released in December 2024. They span six areas:
Recommended Free Tools
- Professionalism
- Patient care and procedural skills
- Medical knowledge
- Practice-based learning and improvement
- Interpersonal and communication skills
- Systems-based practice
These domains offer a useful frame for AI-related learning: the question is not only whether a trainee can operate a tool, but whether they can use it in ways consistent with competent care across these areas. The AAMC also identifies AI among emerging competencies relevant to undergraduate, graduate, and continuing medical education. AAMC’s CBME overview describes the framework and its emerging competency work.
AI literacy is not delegated clinical judgment
AI literacy should help a physician understand a system’s intended use, assess the relevance of its output, communicate about its role, and recognize when to question or set it aside. Delegating judgment is different: it treats an output as a decision rather than as information that must be interpreted in context.
The AAMC’s Principles for the Responsible Use of Artificial Intelligence in and for Medical Education, version 2.0 completed July 31, 2025, put human-centered judgment at the core of responsible use. They also call for transparency, equitable learner access, educator development, interdisciplinary curricula, privacy protections, and frequent evaluation in the setting where a tool is used. These principles are designed to be applied flexibly in local institutional contexts; they are guidance, not a single prescribed curriculum or assessment blueprint.
In a June 10, 2026 policy announcement, the American Medical Association (AMA) likewise said AI should assist rather than replace physician judgment, and called for physician oversight, accountability, transparency, evidence attribution, evaluation, and validation. AMA CEO John Whyte, MD, MPH, said, “AI has enormous potential in healthcare, but it cannot replace physician judgment,” and that “there must always be transparency, accountability, and meaningful physician oversight.” The AMA announcement describes the policies and their context.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What trainees should be able to demonstrate
The following are practical teaching and assessment targets derived from the competency domains and responsible-use principles above; they are not a replacement for local competency standards. They make AI use visible as part of clinical reasoning rather than treating the tool as a shortcut around it.
Frame the task and the tool’s intended use
A trainee should be able to state what question they are asking, what role the AI output is meant to play, and whether the case fits that use. They should distinguish an aid to a particular task from a source of general clinical authority. If the intended use or fit is unclear, the learner should be able to explain why the output cannot settle the decision.
Appraise the output against the case
Assessment should ask learners to explain how an output relates to the patient’s presentation and available evidence, identify information that is missing or inconsistent, and describe what would change their judgment. A fluent or confident answer is not, by itself, evidence that it is correct or applicable. Educators can assess the reasoning process and the learner’s response to uncertainty, rather than scoring only whether the final answer matches an expected response.
Retain core clinical abilities without the tool
For high-stakes competencies, programs should decide which abilities a trainee must demonstrate without AI assistance and which tasks may appropriately include it. That decision belongs to the specific competency and setting. For example, if a learning objective is independent clinical reasoning, an assessment that allows an AI system to supply the reasoning cannot establish that the learner can perform it independently. Conversely, a separate exercise can assess whether the learner uses an approved tool appropriately. The point is to make the assessment conditions match the ability being claimed.
Free tools Windows power users keep installed
One-click scans. No signup required.
Communicate, disclose, and take responsibility
Learners should be prepared to explain relevant AI involvement to colleagues and, where appropriate to the context, patients; identify the basis for a recommendation; and remain answerable for their clinical decisions. Disclosure practices should be set by institutional policy and applicable professional requirements. Training should not imply that a human sign-off alone guarantees safety: oversight needs to be meaningful, and the use of the tool itself must be evaluated.
How educators can design training and assessment
A useful curriculum treats AI as an interdisciplinary subject woven through medical knowledge, patient care, professionalism, communication, and systems-based practice—not as a standalone technical elective. The AAMC principles specifically recommend educator training and interdisciplinary curriculum development, alongside equitable access and privacy safeguards in educational settings ranging from admissions and classrooms to clinical workplaces, coaching, simulation, assessment, and program evaluation.
- Define the competency first. Specify the observable ability and patient-centered outcome. Decide whether the learner must demonstrate independent performance, appropriate AI-assisted performance, or both.
- Set the permitted conditions. State which tools and uses are allowed for a given learning activity or assessment, how their use should be disclosed, and what information must not be entered. Apply institutional privacy and data-handling rules.
- Teach appraisal, not just operation. Use cases that require learners to assess whether the system fits the task, examine the output in context, recognize uncertainty or mismatch, and explain their reasoning. Include situations in which the tool is unavailable or its output should not be followed.
- Assess the claimed ability directly. Match the assessment method to the competency. If the goal is unaided performance, test without AI; if the goal is safe use, assess the learner’s selection, interpretation, communication, and response to the output. Do not infer a skill from tool access alone.
- Prepare faculty and ensure fair access. Educators need enough preparation to teach and assess AI use consistently. Programs should consider whether learners have equitable access to the tools and instruction required for an activity.
- Evaluate in the actual setting and revisit the rules. Assess tools where they are used, protect privacy, and review whether the curriculum and assessment remain appropriate as tools and institutional practices change. The AAMC recommends frequent evaluation rather than assuming a tool is suitable because it works in another context.
How to judge an AI use case before adopting it
Official guidance emphasizes evaluation and validation but does not establish that any one AI system improves or harms medical training outcomes. Institutions therefore need to judge each use case on its own evidence and setting, rather than treating a general claim about AI as proof of suitability.
| Question | What educators or clinical leaders should establish |
|---|---|
| Intended use | What task is the system meant to support, and does the proposed learning or clinical use match that purpose? |
| Evidence and validation | What evidence supports the intended use, and has the system been evaluated in the setting where it will be used? |
| Explainability and attribution | Can users understand the basis for an output well enough to appraise it, and can relevant evidence be identified? |
| Equity and bias | How will the institution assess whether performance or access differs across relevant groups, and what response is planned if inequities are found? |
| Privacy | What data may be entered, how are they protected, and do the proposed practices comply with institutional requirements? |
| Access and education | Can learners and educators access appropriate training and tools without creating unfair differences in learning opportunities? |
| Monitoring | Who will review performance and suitability after deployment, and how will concerns lead to changes in use? |
These are evaluation axes reflected in AAMC, AMA, and Irish regulator guidance, not a head-to-head ranking of products. Human oversight is important, but it does not by itself establish that an AI use is safe or effective.
Professional accountability depends on jurisdiction
In Ireland, the Medical Council’s October 21, 2025 position statement places doctors’ knowledge and expertise at the forefront of clinical decisions. It identifies professional accountability, clinical judgment, transparency, shared decision-making, equity, confidentiality, and education and competence as areas of importance. It also states that existing professional and legal standards apply when AI contributes to errors or adverse outcomes. That is guidance from Ireland’s medical regulator; it should not be presented as a universal legal rule for other jurisdictions.
Medical Council of Ireland CEO Dr Maria O’Kane said, “As AI increasingly makes its way into healthcare, we as the regulator have a responsibility to doctors and patients to guide its safe, ethical, and effective use in clinical decision-making.” The Council’s statement provides the Irish context. In the United States, the AAMC and AMA materials cited above are educational principles and policy positions; educators and clinicians should also follow applicable local law, institutional policy, and professional standards.
The competency bar should rise with the tool’s influence
The more an AI output can shape a learner’s reasoning or a clinical decision, the more important it is to make its role, limits, evaluation, and accountability explicit. Training should produce physicians who can use tools thoughtfully and still demonstrate the competencies their patients depend on when the tool cannot be trusted or is not there.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
