Poster ID
P-32
Poster Title
Trustworthy AI development Using Biobank Data: From Norms to Practices in South Korea
Authors
1. HyeonJeong Park, 2. Minjeong Kwon, 3. Jung Hyun Lee, 4. Wonhoo Yoo, 5. Hannah Kim*
1-4. Asian Institute of Bioethics and Health Law, Yonsei University, Seoul, Republic of Korea
5. Asian Institute of Bioethics and Health Law, Yonsei University, Seoul, Republic of Korea; College of Medicine, Division of Medical Humanities and Social Science, Yonsei University, Seoul, Republic of Korea
1-4. Asian Institute of Bioethics and Health Law, Yonsei University, Seoul, Republic of Korea
5. Asian Institute of Bioethics and Health Law, Yonsei University, Seoul, Republic of Korea; College of Medicine, Division of Medical Humanities and Social Science, Yonsei University, Seoul, Republic of Korea
Abstract
Background
South Korea's 「Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust」, effective January 22, 2026, provides, under Article 27, a legal basis for addressing ethical issues arising from AI technologies. Within healthcare, this builds on a sequence of research ethics guidelines:「Research Ethics Guidelines for Healthcare AI Researchers」 published in 2023, and the 「Research Ethics Guidelines and Checklist for Generative AI in Healthcare」 published this year. However, these guidelines do not fully address the governance needs of foundation models trained on pre-established cohort and genomic data. Their large-scale pretraining and reuse across downstream tasks raise questions about data use beyond the original purpose of collection. Genomic data also require particular safeguards because they may reveal information about both individuals and their biological relatives, requiring a careful balance between public benefit and privacy risk. In response, this study is developing an ethical guideline for foundation model research using pre-established cohort and genomic data.
Methods
We are developing this guideline through a comparative analysis of domestic and international laws, regulations, and guidance. We are also reviewing relevant examples reported in the literature to identify ethical issues arising in AI model development and the use of existing data. Additionally, we are constructing operational principles and data governance considerations specific to pre-established cohort data use, referencing the GA4GH Model Data Access Agreement (DAA).
Conclusion
This guideline aims to give researchers, developers, and Institutional Review Boards a common, practical ethical standard for foundation model research on public genomic and health-related data. Alongside this, the guideline is meant to open space for public deliberation, building the institutional foundation needed for experts and the public to trust how this data is used. In this context, this study identifies the ethical principles and data governance considerations needed to responsibly use public genomic and health-related cohort data in AI research.
Acknowledgement:
This research is supported by the 2026 ‘Establishing Ethical Guidelines for the Development and Use of Healthcare AI Models (2026-ER0902-00)’ funded by the Korea National Institute of Health, the Korea Disease Control and Prevention Agency.
South Korea's 「Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust」, effective January 22, 2026, provides, under Article 27, a legal basis for addressing ethical issues arising from AI technologies. Within healthcare, this builds on a sequence of research ethics guidelines:「Research Ethics Guidelines for Healthcare AI Researchers」 published in 2023, and the 「Research Ethics Guidelines and Checklist for Generative AI in Healthcare」 published this year. However, these guidelines do not fully address the governance needs of foundation models trained on pre-established cohort and genomic data. Their large-scale pretraining and reuse across downstream tasks raise questions about data use beyond the original purpose of collection. Genomic data also require particular safeguards because they may reveal information about both individuals and their biological relatives, requiring a careful balance between public benefit and privacy risk. In response, this study is developing an ethical guideline for foundation model research using pre-established cohort and genomic data.
Methods
We are developing this guideline through a comparative analysis of domestic and international laws, regulations, and guidance. We are also reviewing relevant examples reported in the literature to identify ethical issues arising in AI model development and the use of existing data. Additionally, we are constructing operational principles and data governance considerations specific to pre-established cohort data use, referencing the GA4GH Model Data Access Agreement (DAA).
Conclusion
This guideline aims to give researchers, developers, and Institutional Review Boards a common, practical ethical standard for foundation model research on public genomic and health-related data. Alongside this, the guideline is meant to open space for public deliberation, building the institutional foundation needed for experts and the public to trust how this data is used. In this context, this study identifies the ethical principles and data governance considerations needed to responsibly use public genomic and health-related cohort data in AI research.
Acknowledgement:
This research is supported by the 2026 ‘Establishing Ethical Guidelines for the Development and Use of Healthcare AI Models (2026-ER0902-00)’ funded by the Korea National Institute of Health, the Korea Disease Control and Prevention Agency.
Digital Poster