The increasing adoption of artificial intelligence (AI) by militaries worldwide has intensified, as demonstrated by recent research in China leveraging Meta’s open-source Llama model. Chinese researchers utilized an earlier version of the Llama 13B model, incorporating military-specific parameters to develop a specialized AI tool named “ChatBIT.” Designed for the military sector, ChatBIT aims to support intelligence gathering, provide reliable information for decision-making, and enhance question-answering tasks in operational settings.
ChatBIT: Military-AI Collaboration in China
Chinese researchers reportedly fine-tuned Meta’s Llama 13B model, tailoring it for military applications. According to a paper on the project, ChatBIT has been optimized for intelligence tasks, ranging from data processing to decision-support for strategic operations.
Meta’s Open-Source Policy and Ethical Concerns
Meta’s approach to open-source AI has been relatively liberal, with the company releasing models like Llama for public use. However, Meta’s licensing terms prohibit their AI models’ use in activities related to warfare, nuclear industries, espionage, and other high-risk applications. In response to questions about ChatBIT’s use by PLA-affiliated researchers, Meta emphasized its acceptable use policy, stating that any use of their models by the PLA is unauthorized. Yet, due to the open-source nature of Llama, Meta faces limitations in enforcing these restrictions.
Implications for U.S. National Security
The adaptation of Meta’s AI models by Chinese researchers has intensified debates in the U.S. regarding the national security risks associated with open-source AI. President Joe Biden’s administration, recognizing both the benefits and risks of AI, signed an executive order in 2023 aimed at regulating AI developments. This order sought to balance innovation with the need for security safeguards, as open-source models may be susceptible to misuse by state actors.
China’s Broader AI Ambitions: A Race to Close the Technology Gap
ChatBIT represents just one facet of China’s wider AI ambitions. The country has invested heavily in developing indigenous AI technologies, establishing numerous research labs and cultivating partnerships between civilian and military institutions. China’s objective to lead in AI by 2030 has spurred a surge in AI-related research, often with dual-use (civilian and military) potential. A 2023 report by Georgetown University’s Center for Security and Emerging Technology identified 370 Chinese institutions involved in AI research, underscoring the country’s commitment to AI leadership.
Notably, Chinese researchers have also explored applying Meta’s Llama model to domestic security. For instance, a June paper documented its use in “intelligence policing” to support data analysis and enhance decision-making within law enforcement. PLA Daily, a state-run publication, has even suggested that AI could accelerate the development of new weapons, improve combat simulation, and optimize military training.