Bill Mandating Public Data Provision for AI Training Passes
A bill requiring government agencies to provide public data — created with taxpayers' money — to artificial intelligence (AI) developers upon request has passed the State Council.
The government also approved amendments to the Framework Act on Artificial Intelligence and the Data Industry Act at the same session. The AI framework act amendment cleared the State Council just six months after being introduced in late March. The two amendments now await only presidential promulgation. Once enacted, the barriers to using data held by public institutions are expected to fall significantly.
Why an AI Data Bill Now
As competition in AI technology intensifies, securing training data has emerged as a core element of national competitiveness. AI improves in performance only after learning from vast amounts of data, and the cost and time required to obtain high-quality data have been a heavy burden for companies. In particular, it has been repeatedly pointed out that Korea's relative shortage of language data compared to English-speaking countries has slowed domestic AI development.
Although the government and public institutions have long built public data with taxpayer money, privacy protections and institutional circumstances have limited its actual use in industry. The sense that data exists but remains difficult to access underpins this legislation. Pushing the Data Industry Act amendment alongside it is interpreted as an effort to institutionalize the entire lifecycle of data, from production to distribution.
Key Provisions and Scope
The gist of the amendment is to explicitly require public institutions, in principle, to provide data they hold and manage when requested for AI training purposes. It is also known to include methods and procedures for providing data in refined forms suitable for AI development. Safeguards such as personal information protection obligations remain intact, and the grounds for refusing to provide data are limited by law.
The amendment applies to companies and research institutions seeking to use public data for AI development. Given that small and medium-sized enterprises and startups have comparatively limited data-gathering capabilities, the primary beneficiaries are likely to be developer groups without deep pockets. Public institutions holding the data, on the other hand, face new administrative burdens in setting up provision systems.
Industry Impact and Concerns
Optimism prevails. The bill carries fiscal significance as well, ensuring that data built with taxpayer money is put to use for industrial competitiveness rather than left idle. If securing Korean-language data needed to develop domestic large language models becomes easier, it is also expected to partially ease the reliance on foreign big tech.
Concerns remain, however. Even if public data provision becomes mandatory, the law's effectiveness could suffer if the data is of poor quality or inconsistent in format. Public institutions also note that with narrower grounds for refusal, the potential for disputes will grow. Experts broadly agree that data standardization and safe pseudonymization technology must support the law for its purpose to be realized.
Timeline and Outlook
The two amendments are expected to take effect some time after presidential promulgation. Following promulgation, the specific scope of data provision procedures and refusal grounds will be finalized through the drafting of enforcement decrees and public notices. Since the law's real-world effectiveness will hinge on subordinate legislation, the industry is expected to watch the decree-drafting process closely.
Going forward, the key question will be how to maintain personal information protection standards while opening the door to public data. If only the provision obligation is strengthened while safeguards such as pseudonymization remain inadequate, data subjects' concerns could grow. How many examples of cooperation between public institutions and companies emerge in the early stages of the law's implementation will shape the data foundation of Korea's AI industry.
