This article first appeared in The Edge Malaysia Weekly on September 14, 2026 - September 20, 2026

Artificial intelligence (AI) in medicine is no longer merely a conference topic. It is already entering Malaysian hospitals and clinics — reading scans, flagging the deteriorating patient on the ward, transcribing consultations, suggesting the differential a perhaps tired mind might miss.

The benefit should be stated plainly and honestly, because it is real. A system that never tires, and has read more images than any radiologist will see in a lifetime, is not just a gimmick.

Used well, it means earlier detection, faster triage and, not least, time returned to the doctor for the part of medicine no machine performs — the patient. In a system where specialists are stretched and wards are full, that is not a marginal gain at all.

But precisely because the benefit is real, the questions that follow it deserve straight answers. Three of them matter most — for patients, for hospitals and their boards, and for the profession itself.

Begin with the question every doctor asks first. If an algorithm misses the tumour or recommends the wrong dose, who is liable?

The law’s answer is reassuringly old-fashioned. The algorithm is a tool, not a colleague. Our courts measure a doctor against the standards of the profession, and on that measure, blindly following an algorithm is no more defensible than unthinkingly following a textbook.

The machine proposes; the doctor disposes. Clinical judgment remains sovereign. And because a clinical decision has always had to be capable of justification, a recommendation nobody can explain is one the doctor must treat with particular care, not particular deference.

That cuts in both directions. A doctor who overrides the algorithm for sound clinical reasons, and records those reasons, is practising medicine, and the law will treat him accordingly.

A doctor who accepts its output without applying his own mind has not been replaced by the machine; he has simply stopped doing his job.

Nor does the analysis change because the tool is clever. The junior doctor relies on the senior, the surgeon on the anaesthetist, the clinician on the laboratory — medicine has always been practised on trusted inputs, and the law has always asked the same question of each professional. Did you exercise the judgment the circumstances required?

And the question is beginning to cut the other way. The standard of care is never static; it evolves with what responsible practice regards as standard. As these tools prove themselves and become commonplace, a day may come when the doctor who declines to use a reliable one, and misses what it would have caught, is the one who has to explain himself.

Liability, in short, does not simply transfer to the software.

The duty of care stays where it has always been, with the humans and institutions that use the tool, which is why the next question is the one boards should be asking.

Malaysian law already treats software as capable of being a medical device. Section 2 of the Medical Device Act 2012 says so expressly. Approval to be placed on the market, however, is a floor, not a warrant that a tool is safe in a particular hospital, on a particular population, in particular hands.

The gap between market approval and safe bedside use is the institution’s to close. And the discipline required is one hospitals already know, because it is how they treat people.

A new algorithm should be brought in the way a new consultant is, namely, credentialed before it starts, given a defined scope of practice, supervised while it proves itself, its performance audited against local outcomes, its users trained, and its privileges withdrawn if it underperforms.

An algorithm adopted on a vendor’s brochure and left unsupervised is not innovation. It is an unexamined risk, and when it fails, the questions will be asked not of the software house first, but of the hospital that deployed it and the board that governs the hospital.

Sound AI governance, in other words, is now part of sound corporate governance.

Three questions belong on the agenda of any board whose institution deploys these tools. Who validated this system for our patients, and against what; who is accountable for its performance in service; and how quickly can it be switched off?

If those questions have immediate answers, the institution is governing its tools. If they do not, the tools are governing the institution.

Two other duties complete the picture.

The first is candour. Where an algorithm materially informs a patient’s care, the patient should be told in plain and simple terms, not technical ones. Consent has moved towards the patient’s right to understand what is being done and why; a significant tool in the diagnostic chain is part of that story.

The profession has already moved. The Malaysian Medical Council issued a guideline in 2025 on the ethical use of AI in practice, under which a doctor should, where appropriate, explain to the patient the part the tool plays in their care, what its recommendation rests on and the risks it is known to carry.

It need not be a seminar. In many cases, a plain sentence explaining that a computer system helps analyse the scan and that the doctor makes the final call will go a long way towards honouring the duty, and may reassure rather than alarm.

The second is data. These systems learn from, and run on, the most intimate information a person ever surrenders. The confidence a patient reposes in the consultation must follow the data wherever it flows — into training sets, dashboards and third-party platforms — because the trust between doctor and patient is the one asset on which every one of these systems ultimately depends.

And that is a matter of law as much as of ethics. Data protection duties do not disappear because the processor is intelligent, though in Malaysia the statutory framework binds the private sector more fully than the public, which makes the profession’s own duty of confidence carry all the more weight.

Which leaves the question the profession itself must answer.

The purpose of these tools is not to make medicine automatic. It is to make doctors better. To catch what fatigue misses, to lift the clerical burden, to return the hours now spent on paperwork to the bedside where they belong.

The doctor who uses AI well will serve patients better than the doctor who refuses it, and better than the machine alone. What must not change is where the ultimate judgment sits.

Medicine is practised in the space between the data and the person. In the history that does not fit the pattern, the patient who does not want what the guideline assumes, the decision that weighs a life rather than a probability.

No algorithm occupies that space. The skill to cultivate in the next generation of doctors is not deference to the machine, but discernment with it.

The algorithm has entered the consulting room, and it should be welcomed there. But the duty in that room, to the patient, in law and in conscience, remains, and must remain for all intents and purposes, human.

Professor Datuk Dr Hanafiah Harunarashid is the Master of the Academy of Medicine Malaysia and the chief medical director of KPJ Healthcare Bhd. J J Chan is a barrister-at-law of the Honourable Society of Gray’s Inn, London, an advocate and solicitor of the High Court of Malaya and an adjunct professor of the Faculty of Law, Universiti Malaya.

Save by subscribing to us for your print and/or digital copy.

P/S: The Edge is also available on Apple's App Store and Android's Google Play.