Asei Ito (Associate Professor, Institute of Social Science, the University of Tokyo; HYI Visiting Scholar, 2022-23) and Kota Takaguchi (Visiting professor, Chiba University, and freelance journalist)
Abstract: This study examines how the influence of large language models (LLMs) developed in China is starting to spread beyond the country’s borders. Since 2022, the Chinese government has accelerated its promotion of generative artificial intelligence (AI) technology while simultaneously strengthening regulation and censorship. By 2024, Chinese authorities had established a clear legal regulatory framework for LLMs, which included technical requirements and testing schemes. Focusing on the case of the Japanese market, this study conducted an early empirical assessment and observed that the output of some derivative models—Chinese-language LLMs retrained by Japanese companies—has retained Chinese political correctness. Hence, this study highlights the potential for ideologies—socialist core values—to spread across borders through open-source LLMs. The study provides new insights into how emerging technologies can influence political landscapes beyond their countries of origin through open-source channels.
Keywords: large language models; Qwen; censorship; political correctness; China; Japan
About the Working Paper Series:
The Harvard-Yenching Institute is pleased to make available working papers by HYI affiliated scholars on topics in the humanities and social sciences, with special attention to the study of Asian culture. The HYI Working Paper Series welcomes submissions from all HYI-affiliated faculty and fellowship grantees (including graduate students). Scholars are invited to post papers either in English or in an Asian language. To submit a paper, please email strogatz@fas.harvard.edu.
The views expressed in the HYI Working Paper Series are those of the author(s) and do not necessarily reflect those of the Harvard-Yenching Institute. HYI Working Papers have not undergone formal review and approval. Such papers are included in this series to elicit feedback and to encourage discussion. Copyright belongs to the author(s). Papers may be downloaded for personal use only, and may not be cited without the author’s permission.