The National Artificial Intelligence Strategy Committee on Wednesday held a meeting with national AI companies and discussed ways to secure data needed to develop a Korean independent AI foundation model, known as Dokpamo.

LG AI Research, Upstage, SK Telecom and Motif Technologies attended the meeting and conveyed on-the-ground difficulties related to the Dokpamo project.

LG AI Research stressed the need to overhaul regulations to expand data use. It said it would speed up development of a globally competitive Dokpamo, backed by government efforts such as improving the copyright system.

Upstage stressed that building data infrastructure to strengthen Korean-language AI competitiveness is urgent. It cited digitising legal deposit book data at the National Library of Korea, establishing safe harbour rules for text and data mining (TDM), and building datasets centred on experts' thought processes.

SK Telecom proposed improving verification and refinement systems for shared data provided by the government. It also suggested securing domain-specific data, expanding Korean-language evaluation datasets and resolving public data licensing issues.

Motif Technologies pointed to the reality that it is difficult for individual companies to independently build large-scale Korean-language pretraining data. It proposed building and operating a high-quality national Korean-language pretraining dataset. It requested expanding support for data post-processing and processing in existing support programmes that operate with a focus on securing source data.

Lim Moon-young (임문영), the committee's standing vice chair, said: "The data 확보 issue is a nationwide task, not the role of a specific ministry or institution." He added: "The committee will take the lead in bringing capabilities together and providing swift, effective support."

The committee approved the "Republic of Korea Artificial Intelligence Action Plan" at its second plenary meeting on Feb. 25. The plan includes many tasks related to securing and using data, including securing a general-purpose AI foundation model, upgrading data quality based on the National Data Integration Platform, energising an AI data-sharing ecosystem and overhauling regulations on the use of personal and industrial data, and energising an ecosystem for using copyrighted works for AI training purposes.