AI Regulation Tracker  /  Draft rule

Indonesia is reportedly drafting a "Good AI Medical Practice" guideline for hospital AI

Indonesia's health ministry and drug agency are reportedly drafting a guideline that would make medical AI pass clinical testing and safety review before hospitals deploy it. Diagnostic-AI and clinical-decision-support vendors, including foreign ones, would face local validation.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published July 7, 2026  ·  Last updated July 7, 2026

Indonesia is reportedly preparing to put a gate in front of the medical algorithms its hospitals increasingly rely on. According to secondary reporting, the Ministry of Health (Kementerian Kesehatan, or Kemenkes) and the Food and Drug Authority (BPOM) are drafting a guideline provisionally described as "Good AI Medical Practice." As proposed, it would require medical AI and algorithms used in hospital settings to pass clinical testing, ethical validation, and safety assessment before they are deployed on patients, backed by certification and continuing oversight.

The effort sits inside a larger plan. AI is a named pillar of the government's Health Digital Transformation Blueprint for 2025 to 2030, and this guideline is the piece meant to govern how clinical AI actually enters practice. A companion project, reportedly led by the national research agency BRIN together with the University of Indonesia, is said to be developing an "AI Explainability Framework" so that clinicians and regulators can understand how a given model reaches its output.

What is verified, and what is not

Two points deserve care. First, this is a draft in preparation, not a rule in force. As of the most recent reporting, in October 2025, officials were described as working on the guideline. No effective date has been announced. Second, no primary Permenkes (ministerial regulation) number has been located, and the account here rests on Indonesian secondary press, including Hukumonline, rather than on an official published text. The direction of travel is clear; the operative document is not yet public.

That distinction matters for anyone building a compliance plan. Duties described in secondary reporting can shift in the drafting and inter-ministerial review that Indonesian regulations pass through before signature. Treat the specifics below as the reported shape of a coming rule, not as settled law.

What the rule would require

As reported, the core obligation is a pre-deployment gate. A medical AI tool, whether a diagnostic model reading images, a clinical-decision-support system flagging risk, or an algorithm embedded in a telemedicine workflow, would have to clear three checks before use: clinical testing to show it performs as claimed, ethical validation, and a safety assessment. Certification and ongoing oversight would follow. In practice this treats clinical AI closer to a regulated medical intervention than to ordinary software, which is why BPOM, the drug and device regulator, is reportedly involved alongside the health ministry.

The involvement of two agencies is worth noting. Kemenkes sets the terms under which hospitals and clinicians practice, while BPOM regulates the products, including software that functions as a medical device. A guideline drafted jointly signals that Indonesia intends to police both the tool and its use in care, rather than leaving a gap between device approval and clinical adoption. That is also why the companion explainability work matters: a safety assessment is hard to complete if regulators cannot see how a model reaches its conclusions.

What it does not do

The proposed guideline is not, on the reporting available, a general AI law, and it is not a data-protection rule. It targets AI used in hospital and clinical care, not consumer health apps or administrative software. It also does not, as described, ban categories of medical AI; it conditions their use on passing validation and safety review. And because it is a draft framework rather than an enacted regulation with stated penalties, there is no published sanction schedule to cite. Professionals should not assume any current legal obligation flows from it yet.

The cross-border angle

For a US reader, the practical reach is on the vendor side. A US or other foreign company selling diagnostic AI or clinical-decision-support software into Indonesian hospitals would, if the guideline is issued as proposed, face local clinical-validation and certification requirements rather than being able to rely solely on approvals earned in its home market. That mirrors a broader pattern in medical-AI oversight worldwide, where national regulators want evidence generated or verified under their own standards before a model touches local patients. Firms with an Indonesian sales pipeline should watch Kemenkes and BPOM publications for the eventual text.

For hospitals and clinicians inside Indonesia, the message is preparatory. The tools are already arriving; the rulebook is being written to catch up. Building a clean record of what each deployed algorithm does, how it was tested, and who is accountable for its outputs is useful regardless of when the guideline lands.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.