The Washington Post on December 1 published a feature story titled “How China Is Using AI to Extend Censorship and Surveillance,” detailing how China is rapidly expanding the use of artificial intelligence (AI) in online content control, monitoring of ethnic minority groups, and judicial procedures, making an already centralized system of governance faster, more extensive, and harder to detect. The report cites analysis from Kai-Shen Huang, Director of the Democratic Governance Program at the Research Institute for Democracy, Society, and Emerging Technology (DSET), who observes that China’s judicial system has experienced a “roller-coaster” trajectory in adopting AI—moving from early over-optimism to a more cautious reassessment. Under policy pressure and procurement competition, he notes, the maturity of AI technologies often falls short of their official claims.
The report, drawing on research by the Australian Strategic Policy Institute (ASPI), notes that the Chinese government is actively promoting the use of AI to enhance online censorship. Major technology companies such as Baidu, Tencent, and ByteDance have been enlisted to develop content-filtering and risk-detection tools. AI models can efficiently process massive volumes of posts, flag politically sensitive terms, and reduce the visibility of targeted content, significantly increasing the speed and precision of political censorship. Although AI can automate large portions of initial screening, ASPI notes that human reviewers remain essential—particularly in interpreting coded language and political innuendo—resulting in a hybrid “AI + human” model of information control.
The article also highlights Beijing’s investment in large language models for Tibetan, Uyghur, and Mongolian, aimed at overcoming linguistic barriers and enhancing surveillance capabilities over ethnic minority populations. Local governments and official research institutions have launched related initiatives, expanding AI’s role in societal monitoring. These language models are positioned as tools for opinion analysis and strengthening “ethnic work,” though their future uses remain a subject of international scrutiny.
In the judicial sector, the Post reports that some Chinese courts have adopted AI for case management, document retrieval, and sentencing recommendations. Certain prisons have even deployed AI to analyze inmates’ emotional states. Researchers warn that the decision-making logic and underlying data of such systems are often opaque, making it difficult for defendants to understand how AI-generated conclusions were reached—and nearly impossible to challenge them—raising concerns about transparency and fairness.
Regarding China’s judicial AI, Huang notes that Chinese judicial agencies initially held highly optimistic expectations for AI and hoped it could quickly streamline legal processes. In recent years, however, they have shifted into a more cautious evaluation phase. He explains that strong central policy directives require local officials to demonstrate progress in AI adoption, while companies and research teams, eager to secure government procurement contracts, often overstate the capabilities of their AI systems. The actual performance of these systems frequently falls short, leaving the public and the international community with an incomplete understanding of China’s true AI capacity.
