South Korea launches first government-wide survey of AI training data
South Korea's Ministry of Science and ICT will begin cataloging artificial intelligence training data held across the public sector starting April 10. The survey marks the first full inventory of data scattered among ministries and public agencies, addressing a fragmentation problem that has made it difficult for companies and researchers to locate and use government datasets for AI development.
The National Information Society Agency will conduct the cross-government assessment under authority granted by the AI Basic Act. Officials will evaluate data assets across all ministries and identify 100 datasets with the highest potential for AI training use.
What the survey covers
The inventory goes beyond simply cataloging existing data. It includes datasets that could become usable after processing-such as de-identification or quality improvements-giving a fuller picture of available resources.
Survey items focus on factors directly tied to AI training usability: data type and structure, original purpose, and scope of possible provision. This practical focus reflects the challenge facing government agencies and private AI companies, which struggle to connect fragmented public datasets for training purposes.
How data will be shared
The 100 selected datasets will be provided through an integrated AI training data system currently under development. The ministry is upgrading its existing AI Hub platform to serve this function.
Datasets suitable for online sharing will be made available directly. Data that cannot be publicly disclosed will be provided through "Data Safe Zones"-secure facilities with physical and technical security measures that allow researchers to analyze sensitive information without exposing it. South Korea currently operates 14 of these zones across 11 institutions.
Broader goals
The ministry plans to create a system that guides users from data discovery through acquisition to actual use. Officials are also developing a framework to support trading in AI training data among organizations.
Kim Kyung-man, the ministry's director-general for AI Policy, said the effort reflects a basic principle: "The core of AI performance and quality lies in usable data."
The survey responds to rising demand for training data across industry and government as generative AI adoption accelerates. By systematizing access to public data, the government aims to reduce duplication of effort and create a foundation for more efficient AI development across sectors.
Learn more about AI for Government and Data Analysis to understand how these initiatives affect your organization's AI capabilities.
Your membership also unlocks:
