The American College of Physicians (ACP) has published a 2026 position paper setting ethical guardrails for AI use in medicine — an update to its 2024 guidance. With AI adoption accelerating and no consensus on issues like patient privacy, disclosure, and equitable use, ACP says both physicians and patients need a clear framework to ensure these tools are used responsibly.

The paper frames AI as "augmented intelligence" — a tool to assist, not replace, physicians. It introduces three ethical guideposts: relationality (AI should strengthen, not weaken, the patient-physician bond), self-governance (physicians must maintain independent clinical judgment), and competence (expert, unbiased, patient-centered care). ACP president Dr. Jan Carney also flagged a growing concern: AI "confabulation," where outputs sound convincing but are factually wrong — with error rates exceeding 90% for complex medical literature tasks.

Key Takeaways:

• AI can boost diagnostic accuracy in areas like skin cancer detection and mammogram interpretation
• "Prompt literacy" — knowing how to query AI to minimize errors — is a must-have skill for physicians at all training levels
• Physicians should always verify AI-generated references and weigh outputs against their own clinical judgment
• ACP's four positions emphasize protecting the patient-physician relationship, physician education, clinical independence, and mitigating algorithmic bias

Why it matters: As AI tools embed themselves deeper into clinical workflows, clear ethical standards aren't just nice to have — they're essential to patient safety and trust.