Taiwan Tightens Focus on AI Devices, Streamlines Filings for Low-Risk Products

Taiwan’s Food and Drug Administration has advanced two notable regulatory updates this month. Registrants filing low-risk Class I devices may now use the agency’s Medical Device Premarket E-submission System, which eliminates the need for hard-copy documentation. In addition, the TFDA issued new Good Machine Learning Practice principles on June 8, adopting the ten guiding principles established by the International Medical Device Regulators Forum for the development of AI-enabled medical devices.

Although non-binding, the guidance outlines clear expectations: well-defined intended use, disciplined software and quality engineering, clinical datasets that are representative of their intended populations, risk management that accounts for human factors, robust clinical validation, and continued monitoring following market entry. The principles draw on established international standards and reference prior guidance from the U.S. FDA, Health Canada, and the UK’s MHRA. Manufacturers already aligned with the International Medical Device Regulators Forum’s (IMDRF) framework are likely to find the requirements familiar.

This means that manufacturers relying on clinical data collected outside Taiwan should be prepared to demonstrate that such data is applicable to Taiwan’s patient population. This may require additional validation prior to market entry. The TFDA further expects AI governance to be integrated into an organization’s existing quality management system rather than managed as a standalone effort. It also anticipates ongoing monitoring for model degradation, along with structured management of software updates and retraining after approval.