A South Korean government initiative aimed at developing a nationally independent artificial intelligence system has encountered significant challenges, with reports indicating that more than half of the finalists were found to have utilized foreign technology in their submissions.
South Korea's Ambitious AI Investment
South Korea is making a substantial commitment to artificial intelligence, evidenced by a massive 727.9 trillion won ($495.8 billion) budget approved for 2026. President Lee Jae Myung has more than tripled AI-related investment to 10.1 trillion won ($6.9 billion), viewing AI as fundamental to the country's future economic competitiveness.
However, a controversy surrounding the government's flagship AI competition highlights the difficulties in achieving the technological independence Seoul desires. This situation arises precisely as South Korea dedicates record funding to foster homegrown AI capabilities, yet the companies selected to spearhead this effort are accused of relying on external technologies.
The Goal of Domestic AI Development
The competition, which commenced last June, has the objective of building an AI model exclusively using Korean-developed technology. The underlying aim is to decrease the nation's reliance on the dominant American and Chinese artificial intelligence systems that currently shape the global market.
Achieving Independence Proves Challenging
Of the five companies that reached the finals of this three-year contest, three have been identified as having incorporated at least some computer code from AI systems developed in other countries, including China.
The involved companies argue that it is impractical to disregard existing AI technologies and start development entirely from scratch. Conversely, other observers contend that the integration of foreign tools introduces security risks and undermines the primary objective of creating a genuinely Korean AI system.
Gu-Yeon Wei, an electrical engineering professor at Harvard University familiar with the Korean contest, stated that requiring every line of code to be developed entirely in-house is not a realistic expectation.
“To forgo open-source software, you’re leaving on the table this huge amount of benefit,” Wei remarked.
Nations worldwide are actively pursuing reduced dependence on foreign technology as they strive to build their own AI capabilities. This technology holds the potential for significant impacts on both economic strength and national defense.
South Korea has entered this global race with notable vigor, leveraging its major computer chip manufacturers, software companies, and strong government backing for what officials refer to as sovereign AI.
Competition Aims to Identify Domestic Winners
The competition is designed to select two domestic winners by 2027. These winners must demonstrate performance levels that are at least 95% as effective as the leading AI models from companies like OpenAI or Google. The chosen winners will receive government funding for data collection and personnel, along with access to computer chips procured by the government, which are essential for AI development.
Controversy Surrounds Finalist Submissions
The controversy gained significant attention recently concerning Upstage, one of the five finalists. Ko Suk-hyun, the head of Sionic AI, a competing Korean company, alleged that parts of Upstage's AI model bore resemblances to an open-source system from Zhipu AI, a Chinese company. He also pointed out that copyright notices from Zhipu AI were still present in some of Upstage's computer code.
“It’s deeply regrettable that a model suspected to be a fine-tuned copy of a Chinese model was submitted to a project funded by taxpayers’ money,” Ko stated on LinkedIn. Sionic AI had participated in the South Korean contest but did not advance to the final round.
Upstage responded by conducting a live verification session, presenting its development records to demonstrate that the model was built and trained from its inception using proprietary techniques. However, the company did admit that the inference code, which enables the model to function, did incorporate open-source elements sourced from Zhipu AI, a common practice among developers globally. Ko subsequently issued an apology.
Following this, scrutiny shifted to other finalists. Naver's AI system faced accusations of similarities to products from China's Alibaba and OpenAI in its visual and audio encoders, which are responsible for converting images and sounds into formats that machines can process.
SK Telecom encountered comparable questions regarding its inference codes, which were noted to be similar to those developed by DeepSeek, another Chinese company.
Naver acknowledged the use of external encoders, characterizing it as a strategic decision to employ standard technologies. The company emphasized that its core engine, which governs the system's learning and training processes, was developed entirely by its internal team. SK Telecom echoed this sentiment, stressing that the foundational elements of its model were independently created.
The rules of the competition did not explicitly prohibit the use of open-source code from foreign companies. South Korea's Ministry of Science, which oversees the contest, has not provided any new directives since the controversy emerged. Minister of Science Bae Kyung-hoon welcomed the ensuing discussions.
“As I watched the technological debates currently stirring our AI industry, I actually saw a bright future for South Korean AI,” Bae shared on social media earlier this month.