Sunny Cheung , Kai-shing Lau
On October 21, the Chinese artificial intelligence (AI) company DeepSeek (深度求索) announced the release of a new tool to converts large text datasets into compact image-based formats: DeepSeek-OCR (DeepSeek, October 21). While not the long-awaited R2 large language model (LLM), the firm’s latest release shows that it is continuing to innovate, even as it moves deeper into the orbit of the Party-state. DeepSeek’s success, however, has brought it to the attention of not just the government in the People’s Republic of China (PRC), but also the military.
A procurement platform run by the People’s Liberation Army (PLA) shows that a number of defense companies have won contracts to develop AI tools for the Chinese armed forces. Across the past six months, dozens of distinct procurement documents have explicitly called for tools based on AI models created by DeepSeek. Although the number of procurement notices for military AI are relatively small (the platform posts around 25,000 notices every day), the focus on DeepSeek in the notices that are publicly available is still significant.
DeepSeek’s adoption by the PLA and, to a lesser extent, in the public security domain, is evident beyond this dataset. Research published by military institutions frequently discusses how DeepSeek and other models could be deployed across a range of application scenarios. In some cases, pilots of these systems are already underway. DeepSeek’s models have also been deployed widely across the public and private sectors over the course of 2025, and have become increasingly aligned with state interests (China Brief, March 28, April 25). Given the PRC’s policy of military-civil fusion, PLA deployment of DeepSeek was only ever a matter of time.
No comments:
Post a Comment