Korea is facing an unexpected public controversy on its path to artificial intelligence self-reliance. According to the Wall Street Journal, the Korean government is making great efforts to support the development of "sovereign AI" in order to reduce its reliance on Chinese and American tech giants. However, recent surveys found that the locally developed large models have a high degree of similarity with open-source models from China and the United States at the core code level, sparking heated discussions about the purity of "domestic large models."

In this AI development competition initiated by the government, three out of the five companies that made it to the final stage were accused of using foreign open-source code. The startup company Upstage was publicly questioned, as its model modules were found to be highly similar to the open-source model of Zhipu AI. Even the original author's copyright notice was retained in the code. Subsequently, South Korean tech giant Naver and SK Telecom were also involved in the controversy, as their models were found to have similarities with Alibaba and DeepSeek's products in the visual encoder and reasoning code, respectively.

Faced with these accusations, the relevant companies explained that it is not realistic to write every line of code from scratch in the current technical environment. Naver and other companies emphasized that although they used some external standard tools, the core engine that determines the model's learning and training is still independently developed. Despite the ongoing disputes, the South Korean government has shown a relatively tolerant attitude. The Minister of Science stated that this intense technological debate actually indicates the vitality and future prospects of South Korea's AI industry. Currently, the official plan continues to evaluate the winning models according to the original schedule, aiming to develop a domestic AI tool with global leading performance by 2027.

Key Points:

  • 🇰🇷 Obstacles to Self-Reliance: In order to ensure technological sovereignty, South Korea launched an AI competition, aiming to develop a local model capable of competing with giants like OpenAI by 2027, but it is caught in a controversy over dependence on foreign code.

  • 💻 Use of Chinese Code: Several companies that made it to the finals, including Upstage, Naver, and SK Telecom, were exposed for incorporating open-source elements from Zhipu AI, Alibaba, or DeepSeek into their models.

  • 🛡️ Debate on Technical Approach: Experts believe that abandoning the benefits of open source is not wise, while opponents worry that excessive reliance on foreign tools could pose security risks and undermine the original intention of "sovereign AI."