Six rules for doctors using AI

Doctors should take a balanced approach to artificial intelligence, avoiding both uncritical enthusiasm and a refusal to use the technology on principle, a medical ethics expert has argued.

Writing in The BMJ, barrister and medical ethicist Daniel Sokol set out six practical rules intended to help clinicians use AI safely and responsibly.

First, doctors should understand the system they are using, including its intended purpose and limitations. They do not need to become technical experts, but they should know enough to recognise when a tool may be unsuitable for a particular clinical situation.

Second, any information produced by AI must be checked rather than accepted automatically. AI systems can overlook important details, present uncertain conclusions with false confidence or generate incorrect information. The greater the potential consequences for a patient, the more carefully the output should be scrutinised.

Third, doctors must continue to exercise their own professional judgement. AI can support a clinical decision, but it cannot replace the clinician responsible for assessing the patient and deciding what action to take.

Fourth, patient confidentiality must be protected. Doctors should not enter identifiable patient information into public or unauthorised AI systems. Removing a patient’s name or NHS number may not be sufficient if other details could still reveal their identity.

Fifth, clinicians should remain accountable for decisions involving AI. Under English law, an AI system cannot itself be sued if its advice contributes to patient harm. Doctors should therefore be able to explain why they accepted or rejected an AI-generated recommendation and retain an appropriate record of that reasoning.

Finally, doctors should be open about material uses of AI. Patients should receive enough information to understand how the technology has contributed to their care, particularly where it has influenced an important diagnosis or treatment decision.

The principles reflect a central reality of AI in healthcare: the technology may become increasingly capable, but it does not remove the doctor’s ethical and professional duties. Used carefully, AI could support faster analysis and better-informed decisions. Used without scrutiny, it risks introducing errors behind a convincing appearance of authority.

The safest approach is neither blind trust nor outright rejection, but informed use with human judgement firmly in control.

Stay informed with the stories shaping health and social care.