Abstract
This chapter examines the development of China’s large language models (LLMs) in 2025, focusing on technological progress, industry trends, military applications, and governance practices. It argues that China is rapidly expanding its appeal and influence within the global AI ecosystem through technological refinement, relatively low costs, and an aggressive open-source strategy. China is transitioning from a follower to a leader in several key areas. This shift is evident in the release of open-source models by companies such as DeepSeek and Zhipu AI whose products match or surpass leading Western models in certain evaluations, demonstrating strong algorithmic performance and reasoning capabilities. In response to U.S. restrictions on advanced GPUs, China has adopted a “software-over-hardware” approach, improving model efficiency, advancing compression techniques, and optimizing for domestic chips to maintain momentum despite hardware constraints.
In the military domain, LLMs have become central to China’s “intelligentized warfare” agenda and its broader military-civil fusion strategy. The PLA employs commercial open-source models alongside specialized military systems for command decision-making, intelligence analysis, wargaming, sensor-data fusion, unmanned-system coordination, cognitive operations, and cross-domain kill-chain support, aiming to accelerate operational cycles and improve decision precision.
However, accompanying China’s technological and market expansion is a highly authoritarian AI governance framework. Through layered regulations and end-to-end monitoring mechanisms, the Chinese government maintains control over the generation, dissemination, and use of AI content, ensuring that all AI-derived information remains within politically acceptable boundaries. This governance system effectively makes AI a tool for ideological management and public-opinion regulation. As China exports its technologies abroad, this model risks externalizing its regulatory standards, censorship logic, and potential surveillance mechanisms to other countries that adopt Chinese LLMs.
Therefore, while China presents open-source accessibility and low costs as the outward appeal of its global AI expansion, the underlying foundation is a tightly controlled governance structure closely tied to national security imperatives. This dual strategy means that China’s influence in global AI competition carries not only economic significance but also profound institutional and geopolitical implications.
KEYWORDS: China LLMs; DeepSeek; AI Governance; Military-Civil Fusion; Geopolitics