Chinese Qwen model powers nearly half of South Korea's AI releases this year
A study finds that 49.3% of AI models released by Korean developers between January and August were built on Alibaba's Qwen, highlighting the country's reliance on Chinese technology.
A private think-tank report released Friday reveals that Chinese AI, led by Alibaba's Qwen, underpins 49.3% of AI model derivatives published by South Korean developers on Hugging Face from January to August. Out of 1,845 models, 864 were Qwen-based, dwarfing U.S. offerings that dropped to 35.9% and Korean home-grown models that accounted for just 14.9%. The rapid rise—from 1.2% in early 2023 to nearly half this year—questions the effectiveness of South Korea's sovereign AI initiative, which has been funded with roughly 200 billion won.
Companies such as LG AI Research and Kakao rank far behind, while Naver Cloud was removed from the program for using Qwen components. Developers cite lower operating costs, a wide range of model sizes, multilingual support and easy customization as key advantages of Qwen, despite security worries about data leaks and licensing risks.
Why it matters
South Korea's heavy reliance on Chinese AI challenges its goal of independent technology development and raises security concerns.
In this story
