On 8 June 2026, the Taiwan Food and Drug Administration (TFDA) announced the “Good Machine Learning Practice (GMLP) for AI/ML-Enabled Medical Devices: Development and Management Principles.”
The publication aims to provide manufacturers with a structured framework for the design, development, validation, and lifecycle management of AI/ML-enabled medical devices.
It also aligns Taiwan’s regulatory approach with international best practices, including those proposed by the International Medical Device Regulators Forum (IMDRF).
Read the full announcement here: https://www.fda.gov.tw/TC/siteListContent.aspx?sid=310&id=50868
Purpose of the Draft Amendment
To provide manufacturers with a structured framework for the design, development, validation, and lifecycle management of AI/ML-enabled medical devices.
To align Taiwan’s regulatory approach with international best practices, including those proposed by the International Medical Device Regulators Forum (IMDRF).
To support regulatory oversight of AI/ML-enabled medical devices in Taiwan.
Scope of Application
The guidance applies to medical devices incorporating artificial intelligence (AI) and machine learning (ML) technologies.
This includes software as a medical device (SaMD), whether locally manufactured or imported.
Key Updates
The guidance outlines development and management principles for AI/ML-enabled medical devices.
Key highlights include:
Total Product Lifecycle (TPLC) approach
Emphasises lifecycle management from data collection and model development to post-market monitoring, ensuring continuous safety and performance.
Data quality and governance
Manufacturers are expected to ensure data integrity, representativeness, and proper dataset separation, including training, validation, and testing sets.
Algorithm transparency and documentation
Clear documentation of model design, development processes, training methodology, and performance evaluation is required to support regulatory review and traceability.
Risk management integration
The guidance encourages a risk-based approach throughout the lifecycle, including evaluation of potential clinical risks, methodical risk, security risk, performance degradation, expected benefits and associated patient risks, as well as uncertainty and variability in device inputs, outputs, and clinical use conditions.
Performance evaluation and validation
The guidance emphasises robust pre-clinical and clinical validation, including metrics such as sensitivity, specificity, and generalizability, as well as a focus on human-AI interactions.
Post-market monitoring and change management
The guidance highlights the importance of continuous monitoring, real-world performance tracking, and controlled updates to AI/ML models.
References
- TFDA Announcement on GMLP for AI/ML Medical Devices (Traditional Chinese only)
- Good Machine Learning Practice (GMLP) for AI/ML-Enabled Medical Devices: Development and Management Principles
- IMDRF: Good Machine Learning Practice for Medical Device Development – Guiding Principles
Implications to Client
This guidance reflects TFDA’s increasing focus on regulatory oversight of AI/ML-enabled medical devices and alignment with global regulatory trends.
Clients developing or registering AI/ML-enabled medical devices in Taiwan should:
- Review internal development processes against GMLP expectations
- Strengthen documentation for data management, algorithm design, and validation
- Prepare for more structured regulatory review requirements and lifecycle oversight
- Assess the applicability of non-local datasets to the domestic population and provide supporting evidence where needed
- Establish processes to identify and manage unintended bias and dataset drift
- Demonstrate generalizable performance across the intended patient population
- Monitor model performance over time
For Inquiries
For inquiries or support regarding medical device regulatory requirements in Taiwan, please contact info@nordpacificmed.com.