Artificial intelligence is becoming an increasingly important part of healthcare, with AI-enabled medical devices moving from innovative applications into clinical practice. As discussed in our previous blog on regulatory compliance of AI-enabled devices in the EU and US, the regulatory challenge is no longer simply about demonstrating that an AI-enabled device works; manufacturers also need to understand how AI is incorporated into the device, how its performance is validated, how changes are controlled, and how safety and performance are monitored throughout the product lifecycle.
But what happens when you look beyond the EU and US? In our webinar ‘Registration of AI Enabled Medical Devices Around the Globe’ last June, we discussed more diverse regulatory landscapes for manufacturers considering expansion into Asia-Pacific. Countries such as South Korea, Japan, Vietnam, Singapore, Australia and Malaysia each have their own medical device regulatory frameworks, market-access pathways and approaches to emerging technologies. At the same time, there are important similarities between these jurisdictions. In this blog, we look at these six important Asia-Pacific markets and consider what manufacturers should keep in mind when planning their regulatory strategy.
South Korea – A rapidly developing market for medical technology
South Korea has established itself as an important market for medical technology and digital health, supported by a highly developed healthcare system and strong technological capabilities. For manufacturers of AI-enabled medical devices, this makes South Korea an attractive market, but also one where the regulatory pathway needs to be understood early in the product-development process.
Regulatory Framework & Guidance: Devices are governed by the Medical Device Act, complemented by the Digital Medical Products Act (2025) for digital medical devices. The Ministry of Food and Drug Safety (MFDS) has issued dedicated guidance on the approval and clinical trial design of AI-based devices, including the world’s first guideline on generative AI medical devices (2025). The AI Basic Act, in force since 2026, adds a horizontal AI layer.
Qualification and Categorization: Software qualifies as a medical device if its malfunction could harm patients or it replaces or significantly influences clinical decisions. Under the AI Basic Act, medical devices are considered high-impact AI.
Approval Process: Devices fall into Classes I–IV. Class I devices are notified, Class II devices certified by the National Institute of Medical Device Safety Information (NIDS) or accredited third parties, and Class III–IV devices approved by MFDS, as are lower-class digital health products.
Clinical Evidence: Diagnostic accuracy (e.g. sensitivity, specificity, AUC) must be demonstrated, generally using Korean patient data. Clinical data may be waived for devices equivalent to approved ones.
AI/ML MD Submission Requirements & Obligations: Dossiers must describe the algorithm and training-data management. High-impact AI requires transparency, human oversight and lifecycle risk management.
GenAI: Because generative AI can produce different outputs for the same input, MFDS expects additional verification. Dossiers must identify the foundation model used and explain how its output is processed. Beyond standard performance metrics, manufacturers must evaluate the reliability of generated text using measures such as BLEU, ROUGE or BERTScore. Risks such as hallucination and bias must be identified and mitigated, and users warned about hallucination and performance drift. Model updates are subject to change management and continuous post-market monitoring.
Japan – A mature regulatory system with emerging AI requirements
Japan represents another major medical device market in the Asia-Pacific region, with a mature regulatory system and a well-established pathway for bringing innovative medical technologies to market. For AI-enabled medical devices, manufacturers need to consider how software and AI functionality interact with Japan's existing medical device requirements. The regulatory strategy may depend not only on the intended purpose and risk profile of the product, but also on the nature of the AI functionality and the evidence supporting its safety and performance.
Regulatory Approach & Guidance: AI-enabled devices fall under the Pharmaceuticals and Medical Devices Act (PMD Act) and are typically regulated as Software as a Medical Device (SaMD). The Ministry of Health, Labour and Welfare (MHLW) sets policy, while the Pharmaceuticals and Medical Devices Agency (PMDA) reviews devices. The MHLW’s DASH for SaMD 2 strategy aims to accelerate SaMD development, and the 2025 AI Promotion Act sets a light-touch framework for AI governance.
AIMD Core Requirements: Foreign manufacturers need a local Marketing Authorization Holder (MAH), and all labelling must be in Japanese. Cybersecurity documentation should follow IMDRF principles.
PMDA pathways: Devices are classified as Class I–IV. As certification standards are generally lacking for AI SaMD, PMDA approval is usually required. Early advice is available via the SaMD One-Stop Consultation Desk.
Adaptive AI Lifecycle Management: Similarly to the US PCCP, the IDATEN scheme allows pre-agreed post-market updates to be reported rather than re-approved. Uptake is still limited, but the PMDA plans to extend it to continuously learning AI.
Clinical Evidence & PMS: Clinical performance data must demonstrate algorithm accuracy for the Japanese population, with strictly independent training and validation datasets. The PMDA also oversees post-market surveillance.
Vietnam – Early stage SaMD regulatory development
Vietnam presents a somewhat different regulatory environment from more mature markets such as Japan or South Korea. The key challenge is understanding how the existing medical device framework applies to software and AI-enabled products and how international regulatory evidence can be used to support market access.
For companies developing a regional market-access strategy, Vietnam may therefore need to be considered not only as an individual market, but also as part of a wider Southeast Asian regulatory strategy.
Regulatory Framework, Registration Pathway & Guidance: Medical devices, including software, are regulated under Decree 98/2021/ND-CP (as amended). There is no SaMD- or AI-specific regulation or guidance, no formal adoption of the IMDRF SaMD framework and no dedicated fast track or sandbox. AI-enabled devices therefore follow the standard pathway for their class: Class A/B devices are declared to the provincial Department of Health, while Class C/D devices require a Ministry of Health (MoH) registration certificate based on an ASEAN CSDT dossier. The Law on Artificial Intelligence, in force since March 2026, adds a horizontal, risk-based AI layer.
Classification: The registrant self-classifies the device into Classes A–D under the AMDD-aligned rules of Circular 05/2022/TT-BYT; no SaMD-specific rules exist.
Role of International Approvals/Evidence: Imported devices must already be marketed in another country. Class C/D devices approved in a reference country (e.g. US, EU, UK, Japan, Australia, Canada, China or Korea) qualify for fast-track registration, with the MoH reviewing only legal documents and auditing technical content post-approval.
Clinical Evidence & PMS: No AI-specific clinical requirements apply; clinical evidence is presented within the CSDT. Registration holders must ensure traceability, keep incident records, report serious incidents to the MoH and manage recalls.
Singapore – A leading hub for AI-enabled medical devices and medical technology
Singapore’s Health Sciences Agency (HSA) is considered a leading authority in SaMD and AI regulation because it integrates international standards, lifecycle-based oversight, AI-specific governance, and real-world innovation mechanisms. Thereby creating one of the most mature and forward-looking regulatory ecosystems globally.
Regulatory Framework, Registration Pathway & Guidance: Devices are regulated under the Health Products Act. An early adopter, HSA has set AI-MD requirements in its software guidelines (now GL-04) since 2020. Revision 4 covers machine learning-enabled devices, aligns with IMDRF Good Machine Learning Practice, strengthens cybersecurity and introduces a Change Management Programme (CMP) for pre-agreed software changes under the manufacturer’s QMS. AIHGle 2.0 (2026) complements this framework. Class B–D devices require registration via an ASEAN CSDT dossier.
Classification: Class A–D, with dedicated HSA guidance on SaMD risk classification and CDSS qualification; Class A devices are exempt from registration.
Role of International Approvals/Evidence: Prior approval from an HSA reference agency (US FDA, Health Canada, EU, Japan or Australia’s TGA) opens abridged, expedited or immediate evaluation routes; devices without such approval undergo full evaluation. Reciprocal reliance with Malaysia’s MDA also allows abridged evaluation of MDA-registered devices.
Clinical Evidence & PMS: AI-MD dossiers must describe the datasets (source, size, labelling, bias), the model, performance metrics such as sensitivity and specificity, and the clinical association of the output, with evidence scaled to the software’s clinical significance. Post-market, companies must report adverse events and FSCAs, monitor real-world performance and submit periodic post-market reports for registered AI-MDs.
Malaysia - An evolving regulatory landscape for digital health
Malaysia has become an increasingly relevant market for digital health in Southeast Asia, and its regulatory framework for software and AI-enabled medical devices is developing quickly. While not yet as comprehensive as Singapore or Australia, Malaysia has moved well beyond a general medical device framework, introducing dedicated SaMD guidance, updated classification rules and a regulatory sandbox for AI-driven technologies. For manufacturers, a key consideration is how existing approvals from recognised authorities, including Singapore’s HSA, can be leveraged to streamline market access. Malaysia may therefore be a logical next step for companies that already hold approval in a reference market.
Regulatory Framework, Registration Pathway & Guidance: The Medical Device Authority (MDA) regulates devices under the Medical Device Act 2012 (Act 737). SaMD guidance (MDA/GD/0028) adopts the IMDRF SaMD definitions, and change management guidance MDA/GD/0072 (2026) covers algorithm, software and cybersecurity changes. Registration requires conformity assessment by a registered Conformity Assessment Body (CAB), a local authorised representative holding an establishment licence and GDPMD certification, and a CSDT submission via MeDC@St. A 2025 regulatory sandbox offers novel AI technologies early market access and regulatory guidance.
Classification: Devices are Class A–D under ASEAN AMDD rules. MDA/GD/0062 (June 2025) clarifies that software supporting diagnosis or treatment decisions in non-critical settings is Class B, while software directly influencing real-time care in urgent or high-risk scenarios is Class C.
Role of International Approvals/Evidence: Devices approved by recognised authorities (e.g. EU notified bodies, US FDA, TGA, Health Canada, Japan or UK MHRA) can use the faster verification route. Under the reliance programme with Singapore, an HSA approval can further expedite registration.
Clinical Evidence & PMS: Clinical evidence is provided in the CSDT and assessed by the CAB. Post-market obligations include mandatory problem reporting, field corrective actions and recalls, plus ASEAN post-market information exchange (MDA/GD/0067).
Australia - Strategic gateway for AI-enabled medical devices
The Australian framework provides an established pathway for software-based medical devices, with One of the most comprehensive SaMD regulatory frameworks in the Asia-Pacific region.
Regulatory Framework, Registration Pathway & Guidance: The TGA regulates devices under the Therapeutic Goods Act 1989 and the Therapeutic Goods (Medical Devices) Regulations 2002, amended in 2019 and 2021 to introduce SaMD-specific rules, including exclusions for low-risk software and exemptions for certain CDSS. Amended Essential Principles cover cybersecurity and software lifecycle (EP 12.1) and software version identification (EP 13B). Dedicated guidance addresses software qualification, CDSS, cybersecurity (2025) and AI/ML-based SaMD (2026). An Australian sponsor must include devices in the ARTG; the TGA is consulting on adaptive-AI change control.
Classification: SaMD is Class I–III depending on the seriousness of the condition and whether the software itself provides the diagnosis or treatment decision or informs a health professional.
Role of International Approvals/Evidence: The TGA accepts evidence from comparable overseas regulators (EU notified bodies, US FDA, Health Canada, MDSAP, Japan, Singapore’s HSA and UK approved bodies) to abridge conformity assessment and support ARTG inclusion.
Clinical Evidence & PMS: AI submissions must describe the model and its training and test data, justify generalisability to the Australian population and address risks such as bias and data drift; synthetic data cannot replace clinical data. Post-market, the TGA expects robust performance monitoring and adverse event reporting.
Choosing your market(s)
Although the regulatory pathways differ between South Korea, Japan, Vietnam, Singapore, Australia and Malaysia, the initial strategic questions are quite similar. As with the EU and US, the first step is to define the intended purpose and performance claims of the AI-enabled device. Manufacturers should be clear about what the device does, who it is intended for, the role of the AI within the overall system, and the conditions under which the device is expected to operate.
- Confirm classification per market: The same software may be Class I–IV in Japan and South Korea, Class A–D in the ASEAN markets and Class I–III in Australia, which drives the pathway, timelines and evidence required.
- Leverage existing approvals: US, EU or other recognized approvals unlock abridged or fast-track routes in Singapore, Malaysia, Vietnam and Australia.
- Check data representativeness: Japan and South Korea generally expect local performance data, and Australia requires justification that training and test data generalise to Australian patients.
- Build one core technical file: IMDRF/ASEAN CSDT formats, IEC 62304, ISO 14971 and Good Machine Learning Practice allow documentation to be reused across submissions.
- Plan change management and PMS early: Update mechanisms (e.g. IDATEN, HSA’s CMP, MDA/GD/0072), local representation and post-market obligations differ per market.
With six different markets under consideration, the question arises: where should you launch first? There is no universal answer. The best starting point will depend on the device's classification, intended purpose, existing approvals, clinical evidence, target patient population, local representation requirements, commercial objectives and the manufacturer's ability to support regulatory and post-market activities in each country. Within the region, a common sequence is to start in Singapore or Australia, leverage an HSA approval for Malaysia and enter Vietnam later or in parallel. Japan and South Korea usually require separate planning because of their local data expectations.
For some manufacturers, an established regulatory approval in the EU or US may provide a useful foundation for entering one or more Asia-Pacific markets. For others, the most commercially attractive market may not necessarily be the one with the simplest regulatory pathway. A well-designed strategy can allow manufacturers to reuse the same core evidence, clinical data, risk-management activities, software documentation and quality-system processes wherever possible, while tailoring the final submission to the expectations of each regulator.
The key is therefore to look at the six markets as part of a connected regulatory strategy rather than as six independent submissions. Early decisions on data collection, clinical validation and change management determine how much evidence can be reused across the region and how quickly a first approval can be followed by others.
Planning to bring your AI-enabled medical device to Asia-Pacific? Qserve’s global registration team supports manufacturers with market prioritization, gap assessments and submission preparation. Contact us to discuss your regulatory strategy or watch the recording of our webinar “Registration of AI Enabled Medical Devices Around the Globe”