The MMC Guideline on AI in Medical Practice, Explained for Doctors
Can Malaysian doctors use AI tools? Yes, and they stay fully accountable. The MMC guideline in plain words: training, consent, patient data, with clause numbers.
The Malaysian Medical Council (MMC) allows doctors to use AI and holds them fully responsible when they do. Its guideline says a practitioner should seek appropriate training before using AI, must not share patient data on an AI system without the patient's consent, must validate what the tool recommends, and remains accountable for the clinical outcome.
Part of the guide: AI in Healthcare in Malaysia: Courses and Rules for Doctors, Nurses and Hospitals (2026)

On this page
- What is the MMC guideline on AI?
- Do doctors need training before they use AI?
- Which AI tools may a doctor use?
- Can doctors use ChatGPT under the MMC guideline?
- What must patients be told?
- Can patient data go into an AI system?
- Who is responsible when the AI is wrong?
- What does it say about over-reliance?
- What else is in section 3?
- What the guideline does not say
- A one-page checklist for your department
- How we teach this
- Sources
- Frequently asked questions
This article summarises the guideline for busy clinicians. It is a reading aid. It does not replace the document, and it is not legal advice. The full text is 16 pages and worth your time: MMC Guideline on the Ethical Use of Artificial Intelligence (AI) in Medical Practice, Version 1/2025.
What is the MMC guideline on AI?
It is the Council's statement of how registered medical practitioners in Malaysia are expected to behave when they use AI tools. The MMC Ethics Committee endorsed it on 29 December 2024 and the Council adopted it on 18 February 2025. It was published on the MMC website later in 2025.
The duties sit in section 3, which has fifteen clauses. The first one sets the tone: the existing Code of Professional Conduct and the Council's other guidelines continue to apply. AI adds duties. It removes none.

Do doctors need training before they use AI?
Yes, that is what the guideline expects. Clause 3(4) says practitioners "should seek appropriate training before using AI tools and/or technology in their clinical practice".
The guideline does not name a course, a provider or a number of hours. It leaves the judgement of what is "appropriate" to the practitioner. In practice that means you should be able to explain how the tool you use produces its output, where it is known to fail, and how you check it.
Which AI tools may a doctor use?
Clauses 3(2) and 3(3) say practitioners should use only reliable tools, and describe a standard of "reasonable confidence". The guideline lists what that confidence rests on: the trustworthiness of the tool, whether it is appropriate for the task, transparency about how it works, the standards it meets, and where liability lies.
The guideline does not approve or ban products by name. It does not mention ChatGPT or any other brand. The test is applied by the doctor, tool by tool and task by task. A general chat assistant may pass that test for drafting a teaching handout and fail it for anything that touches a real patient.

Can doctors use ChatGPT under the MMC guideline?
Yes, with conditions. The guideline names no product, so ChatGPT is neither approved nor banned. It speaks of "AI tools and/or technology" and asks for "reasonable confidence" in them under the five tests of clause 3(3): trustworthiness, appropriateness, transparency, conformity to standards and clarity on liability. A general-purpose chatbot has not been validated for clinical decisions, so it passes these tests for some tasks and fails them for others. The table is our reading of the clauses, not wording from the Council.
| Task | Our reading under the guideline |
|---|---|
| Summarise a published paper you will then read | Allowed in practice: no patient data, you read the source |
| Draft a teaching handout or quiz | Allowed: no patient data, check the facts |
| Reword a patient leaflet from a template | Allowed: clinician checks the content |
| Draft a referral letter with identifiable details | Not without patient consent, clause 3(6)(b); use an approved system |
| Paste a consultation transcript | Not without consent, same clause |
| Ask for a diagnosis or dose for a named patient | Not recommended: tool not validated, you stay accountable under 3(5)(c) |
Your employer's rules apply on top. A hospital may restrict tools or require an approved enterprise system, and government hospitals follow Ministry of Health circulars. No written policy at your institution is a gap to raise with your head of department. For a tool-by-tool comparison see Claude vs ChatGPT for doctors.
What must patients be told?
Clause 3(5)(b) requires the practitioner to inform the patient before the consultation when AI is going to be used, and sets out what that information covers. Clause 3(12) adds a specific case: if the doctor and the AI tool reach different conclusions, the patient should be informed.
If you use AI only for your own preparation, such as summarising a paper, there is no patient encounter to disclose. Once a tool plays a part in a consultation, disclosure comes first.
Can patient data go into an AI system?
Not without consent. Clause 3(6)(b) is one of the shortest sentences in the document: "Patient data must not be shared on any AI system without the patient's consent." Clause 3(6)(c) adds that data should be protected with measures such as encryption and access control.
For everyday practice this leads to a simple working rule:
- No names, IC numbers, addresses, dates of birth or images that identify a patient in a public AI tool.
- De-identify properly. Removing the name is not enough if the case is rare or the details are specific.
- Use institutional systems that your hospital has approved for patient data, under its own rules.
- When in doubt, do not enter it.
Who is responsible when the AI is wrong?
The doctor. Clause 3(5)(c) says the practitioner will "remain responsible and accountable for the patient's clinical outcomes". Clause 3(9) says AI "should complement, not replace clinical judgement". Clause 3(13) says practitioners "must validate AI recommendations and intervene as necessary".
Taken together: a recommendation from a tool has the standing of a suggestion from a textbook you have not yet checked. You verify it against the patient in front of you, and the decision is yours.
What does it say about over-reliance?
Clauses 3(10) and 3(11) ask practitioners to be cautious about relying too heavily on AI and to guard against skills degradation. This matters most in teaching hospitals. A house officer who always asks a chatbot for a differential may never learn to build one. Supervisors should decide where AI is allowed in training and where trainees work without it.
What else is in section 3?
Clauses 3(14) and 3(15) cover reporting complications that arise from AI use and declaring conflicts of interest, for example a financial interest in a tool the doctor recommends.
What the guideline does not say
Summaries of this document often add things that are not in it. To be accurate:
- It does not require a specific certificate or accredited course.
- It does not list approved or prohibited AI products.
- It does not name the Personal Data Protection Act. Your duties under data protection law apply separately.
- It applies to registered medical practitioners. Nurses, pharmacists and allied health professionals have their own regulators, although the principles on consent and data are a sound guide for every clinical profession.
A one-page checklist for your department
- Training. Has every doctor who uses AI at work had training on how the tools fail and how to check them? (3(4))
- Tools. Do you have a short list of tools the department considers reliable, and for which tasks? (3(2)–(3))
- Data. Is it written down that no identifiable patient data goes into a public tool, and which systems are approved? (3(6))
- Disclosure. Do doctors know when and how to tell a patient that AI is involved? (3(5)(b), 3(12))
- Validation. Is there a named human check before any AI output is used in care? (3(13))
- Teaching. Have supervisors agreed where trainees may and may not use AI? (3(10)–(11))
- Incidents. Do staff know how to report a problem that involved an AI tool? (3(14))
How we teach this
winsym.ai runs AI courses for doctors, nurses, medical schools and hospital teams in Malaysia. The clinical sessions are led by two emergency physicians, Prof. Dato' Dr. Rishya Manikam and Datin Dr. Sri Latha Balakrishnan, and each one is mapped to the clauses above. The courses are education about AI. They are not clinical training and they do not carry CPD points at present (see CPD points for AI courses). For the wider picture, read the AI in healthcare guide for Malaysia.
Clause 3(4) on training is the reason many doctors look for an AI course for doctors in Malaysia.
Sources
- Malaysian Medical Council, Guideline on the Ethical Use of Artificial Intelligence (AI) in Medical Practice, Version 1/2025, mmc.gov.my/wp-content/uploads/2025/09/MMC-Guideline-on-The-Ethical-Use-of-Artificial-Intelligence-AI-.pdf, clauses re-read in November 2026 (checked November 2026).
Clause numbers and quotations were checked against Version 1/2025 of the guideline on 11 and 22 October 2026. If the Council issues a new version, the newer text applies. General information, not medical or legal advice.
Frequently asked questions
Is the MMC AI guideline legally binding?
It is a guideline of the Malaysian Medical Council for registered medical practitioners, issued alongside the Code of Professional Conduct, which continues to apply. Use the wording of the guideline itself for any disciplinary or legal question and ask a lawyer when it matters.
Does the MMC guideline name approved AI tools?
No. It sets criteria for 'reasonable confidence' in a tool (trustworthiness, appropriateness, transparency, conformity to standards and clarity on liability) and leaves the judgement to the practitioner.
Does the guideline require a certificate or CPD course in AI?
No. Clause 3(4) says practitioners should seek appropriate training. It names no course, provider or hours, and it does not mention CPD points.
How often will the guideline be updated?
The guideline states it will be reviewed annually until 2026 and every two years after that. Check the MMC website for the latest version.
Does the MMC prohibit ChatGPT?
No. The guideline names no products, approved or prohibited. It sets duties that apply to any AI tool a practitioner uses in clinical practice.
Can I paste a patient's history into ChatGPT if I remove the name?
Removing the name is often not enough, especially for rare conditions or specific details. Clause 3(6)(b) forbids sharing patient data on any AI system without the patient's consent. When in doubt, leave it out and use a system your institution has approved.
Do I need to tell patients I used ChatGPT?
If AI is used in the patient's care, clause 3(5)(b) says to inform the patient before the consultation, where appropriate. If you used it only to prepare a talk or summarise literature, there is no patient encounter to disclose.