China Advances AI Surveillance with Minority Language Systems, Introducing a Tibetan Large Language Model

The Chinese government is intensifying its AI development for ethnic minority languages to enhance state surveillance, as per a report by the Australian Strategic Policy Institute (ASPI) on December 1, 2025. This initiative includes the recent launch of SunshineGLM V1.0, the first Tibetan large language model (LLM) aimed at public opinion analysis across multiple platforms, including text, audio, and video. This model was showcased at Tibet University and trained on approximately 28.8 billion tokens of Tibetan data. The National Key Laboratory of Ethnic Language Intelligent Analysis, supported by China’s Ministry of Education, is a key player in this effort, seeking to maintain national stability and monitor communications in languages like Korean, Uyghur, Tibetan, and Mongolian. The report highlights concerns over human rights implications, emphasizing the growing capacity of Beijing to control and oversee communications among these minority groups, which number millions across the PRC.

Source link