Key Points
- Chinese researchers under the PLA adopted Meta’s open-source Llama model to develop ChatBIT, an AI tool for military intelligence and operations.
- ChatBIT’s performance exceeds some models comparable to OpenAI’s GPT-4.
- The tool is optimized for dialogue in military applications, potentially extending to training, strategic planning, and decision-making.
- The U.S. government is heightening security policies, balancing AI innovation with potential national security risks posed by China’s AI advancements.
Chinese research institutions linked to the People’s Liberation Army (PLA) have leveraged Meta’s publicly available Llama model to create a military-focused AI tool. Researchers from institutions under the PLA’s Academy of Military Science (AMS) and the Beijing Institute of Technology fine-tuned an early version of Meta’s Llama model, known as Llama 2, to develop “ChatBIT,” a tool for military operations and intelligence. ChatBIT aims to support intelligence gathering and operational decision-making, excelling at dialogue and question-answering tasks designed for military use.
ChatBIT’s performance exceeds that of some models comparable to OpenAI’s GPT-4, though specifics on its testing parameters and usage are limited. The AI tool has undergone fine-tuning to optimize it for military tasks and may evolve for broader uses, including strategic planning, simulation training, and command decisions.
Meta’s open-source model release policy permits broad public access, albeit with restrictions on certain applications. “Any use of our models by the People’s Liberation Army is unauthorized and contrary to our acceptable use policy,” Meta’s Public Policy Director Molly Montgomery stated. Nonetheless, Meta’s public model access limits the company’s ability to enforce these provisions.
China’s adoption of Meta’s AI model for potential military uses underscores a broader technological challenge for the United States. President Joe Biden’s recent executive order, aimed at balancing AI innovation with security risks, signals the U.S. government’s awareness of potential threats from open-source AI access. Additionally, U.S. agencies are finalizing rules to restrict domestic investment in high-risk tech sectors within China.
While AI development offers considerable benefits for society, Chinese advancements in AI tools like ChatBIT have raised concerns among U.S. policymakers, as open-source models make enforcing usage restrictions difficult. Georgetown University analyst William Hannas notes that China’s ambitions to lead in AI by 2030 are fueled by increasing collaboration between Chinese and American AI scientists. Chinese researchers have also applied Meta’s Llama model to domestic security and electronic warfare, suggesting potential extensions in surveillance and policing.
