Guidelines by field
AI guidelines in medicine and health care what the field's own bodies ask
Health care was among the first fields to write rules for AI, because a wrong answer can reach a patient's body. The World Health Organization, physicians' and nurses' associations, hospital accreditors and regulators now say much the same thing in different words. This page also covers integrative and naturopathic practice.
Last reviewed 27 October 2026. Every claim links to its source.
At a glance
People stay in charge of health decisions
Patient autonomy is the WHO's first principle: machines inform care, people decide it.
A clinician stays responsible
Nurses remain accountable even when technology fails, and AI does not replace a nurse's judgment, assessment or caring presence.
Tell patients where AI touched their care
Physicians' principles call for transparency, and California now requires a disclaimer on AI-drafted messages to patients unless a licensed person reviewed them.
Check for bias and guard privacy
Tools should work fairly across groups, and patient data must be kept secure.
Validate locally, then keep watching
A tool that worked elsewhere must be checked where it is used and monitored after it goes live.
Who has spoken
- Binding law, regulation, court ruling or enforceable professional rule
- Voluntary guidance, principles or a code an organization chooses to follow
Voluntary 2021; 2024
World Health Organization
Ethics and governance of AI for health; guidance on large multi-modal models
Six guiding principles, from protecting autonomy to fairness, followed by guidance for the chat-style models that now draft notes and answer questions, including a call for independent audits after release.
Voluntary Nov 2023
American Medical Association
Principles for AI development, deployment and use
Calls for transparency, for governance to be in place before adopting generative AI, and for privacy and bias protections. Its physician survey found more than 80% of physicians now use AI at work, about double the share in 2023.
Voluntary 2022; republished May 2025
American Nurses Association
The Ethical Use of AI in Nursing Practice
Nurses remain accountable for care, and AI supports but does not replace their judgment and presence at the bedside.
Voluntary Sep 2025
The Joint Commission and the Coalition for Health AI
Responsible Use of AI in Healthcare
The hospital accreditor's first guidance on AI, with governance playbooks and a voluntary certification to follow. Central idea: validate tools locally and monitor them.
Binding Aug 2025; Jan 2026
US Food and Drug Administration
Guidance on AI-enabled devices; revised decision-support and wellness guidance
Software that counts as a medical device needs clearance by law. By mid-2026 the FDA had authorized about 1,450 AI-enabled devices and, as far as is known, no generative-AI device. In January 2026 it loosened oversight of some decision-support software and wellness wearables.
Voluntary Jan 2026
European Medicines Agency, with the FDA
Guiding principles of good AI practice in drug development
Ten joint principles for AI used across the life of a medicine, building on the EMA's 2024 reflection paper.
Binding In force Jan 2025
State of California
AB 3030
Clinical messages to patients drafted by AI must carry a disclaimer, unless a licensed provider reviewed them. Violations can lead to licence discipline for physicians.
AI does not replace a nurse's decision-making, judgment, critical thinking, or assessment skills.
Binding and voluntary
Binding
FDA clearance for software that counts as a medical device; the EU AI Act's treatment of AI in regulated devices as high-risk (now delayed to August 2028); and state laws such as California's AB 3030.
Voluntary
WHO guidance, the AMA's principles, the ANA's position statement, the Joint Commission and CHAI guidance, and the joint EMA and FDA principles for drug development.
What changed in 2025 and 2026
- The hospital accreditor and the Coalition for Health AI issued their first joint guidance in September 2025.
- The FDA finalized its guidance on planned changes to AI devices in August 2025 and, in January 2026, eased oversight of some decision-support software and wellness wearables.
- Regulators on both sides of the Atlantic issued ten joint principles for AI in drug development in January 2026.
Where people still disagree
How much oversight decision-support software needs is unsettled: the FDA's January 2026 changes loosened it, while hospital accreditors are building more governance, not less.
A note
Integrative and naturopathic practice
No naturopathic professional body or licensing board found in this research has issued guidance on AI, as of October 2026. The American Association of Naturopathic Physicians held a conference panel on AI in practice in July 2026, but has no position statement on it. When an Oregon licensee asked about AI scribes in January 2025, the Oregon Board of Naturopathic Medicine replied that it did not have a statement at this time, and its February 2026 agenda had no AI item. Naturopathic bodies outside the US were not checked, so their status is unknown.
Neighboring professions have written models a naturopathic doctor can learn from. Australia's health practitioner regulator explains how existing duties apply to AI: understand the tool, its training and limits, tell patients, and get consent before an AI scribe records a consultation. The UK's osteopathic regulator issued interim guidance anchored in its practice standards in August 2025. In July 2026 the American Osteopathic Association endorsed OsteopathicAI, a minimum standard built around whole-person care, disclosure of patient-facing AI, and checking AI used for osteopathic methods against osteopathic expertise.
The parallel for integrative practice would be checking AI output on botanicals, interactions and treatment reasoning against one's own training and evidence. That parallel is an inference, not a published standard.
Questions to ask yourself
- If this AI-drafted note or message is wrong, would I catch it before the patient acts on it?
- Have I told the patient, in words they understand, where AI touched their care?
- Was this tool checked on people like my patients, or only somewhere else?
- Did my patient agree, before the visit started, to an AI scribe listening?
- If a tool suggests a botanical protocol, have I checked it against interactions, the patient's medicines and my own training?
How this lines up with my everyday guidelines
| My everyday guidelines | The field's nearest rule |
|---|---|
1. Maintain your expertise Your professional knowledge stays primary. AI is collaborative support. | A clinician stays responsible, and AI does not replace professional judgment (ANA; Ahpra). |
6. Stay curious and critical Check sources and verify claims, especially for health and medical information. | Validate tools locally and monitor them after deployment (Joint Commission and CHAI). |
5. Protect privacy and boundaries Do not share client identities or sensitive personal information. | Guard patient data, and get consent before an AI scribe records (AMA; Ahpra). Asking consent before recording is not part of my everyday guidelines. |