Taiwan scholar warns Beijing may use AI to monopolize Chinese-language interpretation

Hung Tzu-wei, a research fellow at Academia Sinica’s Institute of European and American Studies, has warned that Beijing may attempt to monopolize global Chinese-language interpretive authority through artificial intelligence. During a seminar in Taipei, Hung argued that by supplying vast quantities of Simplified Chinese data to train large language models, China could exert influence over how AI interprets language, culture, and knowledge worldwide.

Hung categorized China’s influence operations into two types: short-term interference, such as election-related campaigns, and long-term cultural infiltration. He noted that while the former draws immediate government attention, the latter poses a deeper threat by potentially shaping the ideologies and values of younger generations. This process was described as a form of cognitive warfare that could lead to “epistemic injustice.”

While Taiwan is developing its own language models, Hung cautioned that “databases nobody uses have no future,” highlighting the challenge of competing with the massive datasets available from mainland China. The concern is echoed by Taiwan’s Ministry of Digital Affairs, which noted that Chinese-language training material used by major international large language models remains predominantly Simplified Chinese.