• ACP recently issued four key position points for thoughtful AI use in healthcare.
• The organization’s president told Healio they are “grounded in the patient-physician relationship.”

In a new position paper, ACP detailed key principles for ethical, responsible AI use in healthcare.

AI — when implemented thoughtfully — has the potential to strengthen clinical decision-making, advance patient care “and preserve the central role of the patient-physician relationship,” according to an ACP press release. However, the technology is being adopted rapidly and “there has not been consensus on guidance around issues” like disclosure, fairness and privacy, and both “physicians and patients need practical ethical guidance to help ensure these technologies are used responsibly, equitably, and in ways that support high-quality, patient-centered care.”

Healio spoke with ACP president Jan K. Carney, MD, MPH, MACP, FRCP, to learn more about ACP’s position, the reasoning behind it, deskilling and more.

Healio: Why did ACP publish this position paper? Why is it important to address ethical AI use in medical practice?

Carney: ACP published this paper to address potential benefits and risks of the rapid acceleration of AI use in healthcare from an ethics and professionalism perspective. This position paper updates one from 2024 that first made recommendations in 10 areas. The current paper considers AI as augmented intelligence, meaning that it is a tool to assist physicians and patients, noting that there are at least 200 guidelines emerging in the literature, but no consensus on key issues such as patient privacy, disclosure of AI use and equitable use of AI to help patients.

Healio: What are the potential benefits and consequences of AI use in medicine?

Carney: AI can help physicians in diagnosing disease, identifying treatment alternatives, and making preventive care recommendations. It is particularly helpful, based on current literature, in helping physicians improve diagnostic accuracy in finding skin cancer and interpreting mammograms.

Healio: What are the key components of ACP’s position paper? What values is this policy grounded in?

Carney: ACP’s paper is grounded in the patient-physician relationship as the foundation for the best possible healthcare outcomes. This ethical foundation addresses the core responsibilities of physicians in the patient-physician relationship, including always doing what is in the patient’s best interests, doing no harm, respecting patient autonomy and, in the setting of limited resources, providing care for as many patients as possible.

Patients trust physicians for their evidence-based knowledge, as well as the judgement and trust that defines the patient-physician relationship, especially over time. This trust comes from ongoing communication, considering the clinical evidence and judgement for each patient, and considering patient values and preferences. Physicians are trusted sources of health information and advice, based on science and evidence, and applied to each patient. This patient-physician relationship along with accountability and trust requires the human elements essential to the practice of medicine, elements never reproducible through chatbots.

Healio: The policy paper mentions a hot topic, deskilling . Is there a way to prevent deskilling and ensure AI is used as a way to augment physician intelligence instead?

Carney: I think about deskilling as letting AI think and reason for you, leading to less rigorous or less intensive critical thinking. Automation bias suggests that something from a machine may be more accurate or reliable. In fact, with AI, you must be cautious because AI has varying rates of hallucinations — what I call confabulation — where the material sounds very convincing but is totally wrong. Rates of hallucinations and confabulation vary by how complicated the task or question is, with rates often 5% or lower for simple questions. For scientific questions or medical literature reviews, rates of confabulation can be over 90% in some studies. Confabulation about clinical guideline questions or tasks requiring clinical reasoning vary widely. Teaching physicians — at all levels of their medical education — how to generate questions that minimize this risk, called “prompt literacy,” is essential. Another good practice is to verify the references when asking a question, and to consider the AI response against your expert judgment as a physician for the individual patient.

Healio: Is this an area where regulation should be considered? Why or why not?

Carney: Rather than calling for specific federal regulations, ACP supports oversight and governance of AI’s use in healthcare, outlining key areas in its 2024 position paper. The most recent 2026 paper outlines three ethical guideposts for AI use in healthcare and medicine. The first is termed relationality, reinforcing the principle that AI supports, not weakens, the patient-physician relationship. Self-governance, the second guidepost, refers to physician integrity and independent clinical judgment. Competence means physicians using their expert medical knowledge to help patients, uniquely for each patient, fairly, equitably and without bias.

Healio: Are there any policies that primary care providers should put in place? Any best practices they should know about?

Carney: PCPs, whether in health systems or independent practice, should develop habits and best practices that minimize documented AI risks, such as bias and confabulation, and maximize potential benefits from AI use, using skills of prompt literacy. ACP’s three ethical guideposts provide a practice framework in the context of the patient-physician relationship.

Healio: What is the take-home message here?

Carney: I would love readers to read the paper! The three ethical guideposts provide a practical framework to think about AI use in clinical practice. One of the recommendations in the paper says it best, “when physicians use AI to support medical decision making, that use should respect and promote — not replace — physician-independent clinical judgment and recommendations based on the patient’s individual needs and values.” AI use is changing rapidly. Our goal is to minimize risks to patient care and safety, and maximize potential benefits to patients and to clinical practice. ACP has also developed an AI Resource Hub.

Healio: Is there anything else you would like to add?

Carney: ACP’s four positions in this paper are particularly helpful, emphasizing the patient-physician relationship, physician education in the appropriate use of AI, AI’s use in supporting, not replacing, physician independent clinical decisions and promoting health equity by mitigating bias in AI use. These are quoted below:

• “Position 1. The patient–physician relationship is the core of healthcare. AI must be implemented clinically to support, not weaken, this relationship.
• Position 2. AI is affecting medical practice, research and learning. Physicians and trainees must recommit to the competencies of caring and develop new competencies on the appropriate use of AI.
• Position 3. When physicians use AI to support medical decision making, that use should respect and promote — not replace — physician-independent clinical judgment and recommendations based on the patient’s individual needs and values.
• Position 4. Physicians, in coordination with clinics, hospitals and health systems, must seek to identify and mitigate the effects of algorithmic bias in the use of AI, including by addressing root social and structural causes.”

Jan K. Carney, MD, MPH, MACP, FRCP, can be reached at jacarney@acponline.org.