South Korea is introducing artificial intelligence into its national research and development (R&D) budget review process for the first time, marking a major shift from the traditional system in which officials manually reviewed thousands of project proposals each year. Under the new approach, AI will assist with identifying overlapping projects and drafting preliminary review opinions, as the government accelerates efforts to build a “paperless” budget review system.
The Ministry of Science and ICT unveiled the new AI-powered budget review service during a press briefing at the Government Sejong Convention Center on Thursday. The system was developed using “Solar,” an open-source AI model created by South Korean AI company Upstage.
The platform was jointly developed by the Ministry of Science and ICT, the Korea Institute of Science and Technology Information (KISTI), and the Electronics and Telecommunications Research Institute (ETRI). Officials said the system was built within a closed network separated from external internet access and has passed security verification by the National Intelligence Service, reflecting concerns over the sensitivity of national budget data.
Government officials described the R&D budget review process as one of the most labor-intensive administrative tasks in the public sector. Every year during the May–June review period, experts and government officials must analyze thousands of pages of project plans and budget requests.
One of the most time-consuming tasks has been identifying overlapping or duplicate projects. Since government funding is involved, reviewers must verify whether similar programs have already been approved. However, with more than 1,000 projects under review, the process often depended heavily on human memory and experience.
The newly introduced AI system is designed to automate much of that work. It was trained on approximately 5,000 government R&D budget proposals, planning reports, and expert review documents accumulated over the past five years. The system is also connected to 12.43 million research outcome records stored in the National Science and Technology Information Service (NTIS) database to improve analytical accuracy.
Unlike conventional keyword-based search systems, the AI analyzes context and research objectives to identify similar projects. When a new proposal is submitted, the system automatically suggests comparable past projects and flags items with a high likelihood of duplication for priority review.
During a live demonstration, the AI automatically extracted projects similar to a newly submitted proposal and generated key review points based on the analysis. Functions for summarizing meeting minutes and drafting review reports were also showcased.
Experts who previously participated in budget reviews said the AI significantly reduced administrative workload. However, concerns were also raised over the possibility of AI hallucinations. Some reviewers noted cases in which the AI produced awkward logical connections or exaggerated the significance of certain issues during contextual analysis. Experts said the system would need further refinement to evaluate policy validity and logical consistency beyond simple information retrieval.
The government views the initiative not merely as a pilot project but as the starting point for building an independent AI-driven administrative system. Officials particularly emphasized that the platform was built on a domestic AI model rather than relying on foreign big-tech systems.
Upstage’s Solar model, which powers the platform, is considered one of the leading candidates for South Korea’s planned “independent AI foundation model” initiative. The government aims to reduce reliance on overseas AI technologies in the public sector while building an AI ecosystem optimized for Korean-language data and domestic policy environments.
Bae Kyung-hoon, Deputy Prime Minister and Minister of Science and ICT, described the project as one of the first cases in which a government ministry has directly applied an independent AI model to real administrative work. He added that the government plans to expand AI support in the future to cover earlier stages of project planning and budget requests across ministries.
The government intends to complete AI integration across the entire budget review process by next year. In the longer term, officials plan to expand AI usage into broader policy decision-making areas, including identifying inter-ministerial collaboration projects and analyzing national R&D investment portfolios.
The initiative is already generating both optimism and caution within government circles. While many expect AI to improve efficiency by reducing repetitive tasks, others warn that excessive reliance on AI-generated judgments could weaken accountability and professional expertise in policymaking.
