Healthcare AI Regulations in Asia: A Country-by-Country Guide

Quick answer: There is no single rulebook. In Asia, healthcare AI is usually regulated as a medical device (India, China, Japan) and, increasingly, under a national AI law as well (South Korea, Vietnam). Most companies need local approval, a local representative or license holder, and separate compliance with data protection laws.

If you build or buy medical AI for Asia, you will probably be told, “just follow the rules.” The trouble is that five major markets have five different sets of rules. Some changed this year. This guide explains what applies in each country and what to do first.

Healthcare AI regulations in Asia at a glance

What is the medical AI regulatory framework in Asia?

A medical AI regulatory framework answers three questions. Is your software a medical device? How risky is it? And what proof do you need before it reaches patients?

Most Asian regulators lean on the same international ideas, such as software as a medical device (SaMD) and risk-based classification. They differ on timing, paperwork and who can sell your product. The country sections below follow that logic.

Healthcare AI regulations in India

India does not have a standalone AI law for healthcare. It regulates AI through two channels.

1. Medical device rules. Software with a medical purpose is a medical device under the Medical Devices Rules, 2017. On July 21, 2026, the Central Drugs Standard Control Organisation (CDSCO) issued its final guidance on medical device software. It explains how software is classified, what documents you need and what happens after launch. The guidance describes current practice rather than adding a new control. Software falls into risk classes A to D. State authorities handle the lower classes, and CDSCO handles the higher ones.

For AI products, the guidance expects evidence that your model works in populations, settings and conditions that match its intended use. A model validated only on one hospital’s data will struggle here.

2. Data protection. The Digital Personal Data Protection Act, 2023, and its rules make hospitals, labs and health-tech vendors responsible for patient data, with penalties up to ₹250 crore. ICMR’s ethical guidelines add expectations on transparency and accountability.

Practical tip: If a clinician would act on your software’s output, treat it as a regulated device. CDSCO also formed an expert panel in September 2026 to look at AI in drug discovery, so expect more guidance.

Healthcare AI regulations in China

China has one of the most mature systems for AI medical devices. The National Medical Products Administration (NMPA) classifies AI software by risk. Software that supports treatment decisions with less mature algorithms is generally Class III, while software used for reference or data processing tends to be Class II. The NMPA’s 2022 technical review principles for AI medical devices set the evidence standard.

Three 2026 developments matter:

•    In April, the NMPA published its “AI + Drug Regulation” implementation opinions, a roadmap that runs through 2030 and 2035.
•    On July 30, China issued GB/T 47941.1-2026, its first national standard for AI medical devices. It covers terminology and takes effect August 1, 2027.
•    The NMPA has approved more Class III AI devices this year, including pathology software, endoscopy systems and surgical navigation, so treatment-side AI now gets device-grade review.

Data law is the other half. The Personal Information Protection Law, Data Security Law and Cybersecurity Law all apply when you train or run models on patient data. Plan data flows before you plan submissions.

Healthcare AI regulations in Japan

Japan regulates AI as SaMD under the Pharmaceuticals and Medical Devices (PMD) Act. The route depends on class:

•    Class I: notification
•    Class II: certification by third-party bodies
•    Class III and IV: direct review by the PMDA

Foreign companies must work through a Japanese marketing authorization holder.

Japan’s best feature is its approach to learning algorithms. The IDATEN system lets you pre-agree planned model changes with the regulator, so you can update later without a fresh application. The government’s DASH for SaMD strategy adds a one-stop consultation desk and faster reviews. By September 2025, the PMDA had approved 51 AI-based products, so the pathway is well used.

On broader AI policy, Japan relies on principles and its AI Promotion Act instead of a strict EU-style law. That makes it comparatively flexible, but your device paperwork still has to be rigorous.

Healthcare AI regulations in South Korea

South Korea added a new layer in 2026. Its AI Basic Act took effect on January 22, 2026, and healthcare is a named “high-impact” domain. High-impact operators must assess whether they qualify, explain AI outcomes meaningfully, manage risks and keep human oversight in place. Users must be told when a product uses AI.

Foreign companies that meet the thresholds must appoint a domestic representative. Fines are capped at KRW 30 million, and the government signalled a grace period through January 2027. Sector-specific rules for healthcare are expected around then.

The AI Basic Act does not replace device approval. The Ministry of Food and Drug Safety (MFDS) still reviews AI-based medical devices, so Korean projects usually need both.

Healthcare AI regulations in Vietnam

Vietnam is the newest regime. Its Law on Artificial Intelligence took effect on March 1, 2026, with a risk-based system of high, medium and low tiers.

High-risk AI covers systems that could seriously harm life or health. Obligations include conformity assessment, risk management, human oversight and registration. Foreign providers of high-risk systems must have a local presence. On June 30, 2026, the Prime Minister issued Decision 33/2026, a list of high-risk systems that took effect August 15, 2026. Only listed systems face the heaviest duties, and providers can be liable for damages even when they follow the rules.

Existing healthcare AI systems got a longer grace period, until September 1, 2027. Don’t treat that as a reason to wait. Detailed implementing rules are still being clarified.

Country-by-country healthcare AI regulatory comparison

Medical AI approval requirements in Asia: a practical checklist

Healthcare AI regulatory compliance for companies comes down to a repeatable routine:

1.    Define the intended use. Regulators classify you by purpose, not by technology.
2.    Classify early. Map your product to each country’s risk class before you design clinical studies.
3.    Plan local validation. Expect to show that performance holds in local populations and settings.
4.    Name your local partner. You may need a licence holder, agent, MAH or representative.
5.    Separate data compliance from device compliance. You can pass one and still fail the other.
6.    Document model change control. Decide how updates will be approved in each market.
7.    Monitor continuously. Laws change quickly: India, China, Korea and Vietnam all moved in 2026.

Frequently asked questions

Does Asia have one healthcare AI law?

No. Each country sets its own rules, and only South Korea and Vietnam have dedicated national AI laws.

Is AI software regulated as a medical device in Asia?

Usually yes, when it has a medical purpose such as diagnosis, monitoring or treatment support.

Which country is easiest for AI approval?

It depends on your product. Japan’s IDATEN is friendly to learning algorithms, but every market requires local evidence and a local partner.

Do foreign companies need a local entity?

Almost always, in some form: a licence holder, agent, MAH, representative or local presence.

What is the biggest 2026 change?

South Korea’s AI Basic Act, Vietnam’s AI Law, India’s final software guidance and China’s first AI medical device standard.

Final thoughts

Healthcare AI regulations in Asia reward companies that plan country by country instead of copying one approval everywhere. Start with classification and local partners, treat data law as its own workstream, and revisit your plan every quarter.

Note: This article is general information, not legal advice, so confirm details with local counsel or a regulatory consultant.