Hong Kong Revises MDACS Standards in RS-01:2026(E)

Hong Kong's Medical Device Division (MDD) has issued RS-01:2026(E), Edition 2, its updated List of Recognised Standards for Medical Devices. The 29 October 2026 notice makes the list effective from that date. The official RS-01 document records the previous edition's revision date as 12 May 2021.

The revision history identifies three changes: an updated standards list, a revised Introduction, and a new Section 3.9, Artificial intelligence. For manufacturers and Local Responsible Persons (LRPs), the operational question is how to select and identify standards in a Medical Device Administrative Control System (MDACS) listing dossier, particularly for software and AI-enabled devices.

RS-01 applies only to MDACS. The GN-01 guidance describes MDACS as voluntary. RS-01 Section 1.1 also states that manufacturers and LRPs must follow other standards needed to comply with applicable Hong Kong legislation. Recognition in this list does not by itself establish compliance with every requirement for a device.

Version selection is tied to checklist compilation

Revised Section 1.1 says relevant harmonized and international standards should be cited with their specific publication year or version. Selected standards should reflect the state of the art at the time the Essential Principles Conformity Checklist is compiled, unless use of a historical version is justified. The section refers to Technical Reference TR-004, Essential Principles of Safety and Performance of Medical Devices.

The same section sets a limit on what a standards claim establishes. Recognised standards may not address every listing requirement, particularly for new device types and emerging technologies. Other supporting documentary evidence should then be submitted for evaluation. Even conformance with recognised standards may be insufficient on its own to demonstrate full compliance with the Essential Principles.

Pure Global analysis: a dossier should connect each cited edition to the safety or performance aspect it supports. Listing a recognised code without its year, applicability and supporting evidence does not answer that question. Where an older edition is used, prepare the justification rather than assuming that inclusion somewhere in the catalogue resolves the version decision.

Twelve standards in the new AI section

Section 3.9, on numbered pages 14–15, lists two ISO/IEC standards and ten Mainland China YY/T medical-device standards. The descriptions below summarize their subject matter; the English descriptions of YY/T standards are translations of the Chinese titles printed in RS-01, not quotations from the standards themselves.

These entries establish recognition and identify the listed subject matter. RS-01 does not reproduce the standards' detailed test methods or acceptance criteria; those require the relevant authoritative standard text.

Read AI recognition alongside software and other standards

Section 3.8 separately lists software life-cycle and risk references, including IEC 62304:2006+AMD1:2015 CSV and IEC TR 80002-1:2009. It also lists ISO/IEC 27001:2022, its 2024 amendment, ISO/IEC 27032:2023 and IEC 81001-5-1:2021 for information security and health-software cybersecurity. These are entries in the current edition; the revision history does not individually identify which software entries were newly added.

The catalogue extends beyond software. Examples include ISO 10993-1:2025 in Section 2.1 on biological evaluation, ISO 14155:2026 in Section 2.2 on clinical investigation, ISO 13485:2016 in Section 2.3 on quality management, ISO 14971:2019 in Section 2.4 on risk management, and ISO 17665:2024 in Section 3.10 on sterilization. These examples help locate relevant entries; they do not mean every listed standard applies to every product.

What should manufacturers and LRPs change?

Pure Global analysis: use the revised Introduction and new AI section to review the evidence map for a listing application:

1. Check exact editions when compiling the checklist. Record the publication year or version and assess the state-of-the-art position at compilation. Document the reason for retaining a historical edition.
2. Select AI standards by intended function and evidence need. A dataset or algorithm-performance standard for one imaging application is not automatically applicable to another. Separate standards relevant to terminology, datasets, traceability and the specific clinical application.
3. Map existing GB/YY evidence to the recognised entry. Where an existing report concerns a listed standard, check its edition, scope and covered Essential Principles. Recognition alone does not show that a particular NMPA submission or test report will be accepted by MDD.
4. Keep evidence gaps visible. Review software, cybersecurity and AI aspects together where relevant, and identify requirements that standards do not address. Section 1.1 allows supporting documentary evidence to be evaluated; it does not promise automatic acceptance.

A separate 29 October notice on Guidance Notes and Codes of Practice says applications already submitted through the Medical Device Information System need not be resubmitted because of those GN/COP changes. That assurance belongs to the companion changes; it should not be read as a general exemption from the RS-01 standards-selection expectations.

See our Hong Kong market page for the broader listing context.