Chinese military research institutions linked to the People's Liberation Army (PLA) have reportedly adapted Meta's open-source Llama model to create an AI tool, “ChatBIT,” for potential military applications, according to academic papers reviewed by Reuters. The study, authored by six researchers from the PLA’s Academy of Military Science (AMS) and other institutions, reveals how ChatBIT was developed to support intelligence gathering and operational decision-making.

Using an early version of Meta’s Llama 2 model, the researchers integrated unique parameters to design ChatBIT, which they claim is optimized for military-related dialogue and question-answering tasks. According to the researchers, ChatBIT outperformed similar models, with an efficiency reported at 90% of OpenAI’s ChatGPT-4 capabilities. The exact performance measures and its operational status remain undisclosed.

Meta's AI models, while openly accessible, are restricted under policies prohibiting their use in warfare, espionage, or military activities. However, the open nature of Llama’s release limits Meta’s enforcement options. Meta’s director of public policy, Molly Montgomery, asserted that PLA use of the model is unauthorized and counters Meta’s policies, emphasizing that “the global AI competition” compels the U.S. to keep pace with China’s extensive AI investments.

The PLA Daily and other state media have advocated using AI to support military advancements, including combat simulations and training. Observers suggest China’s AI initiatives are narrowing its technology gap with the United States. William Hannas of Georgetown University’s Center for Security and Emerging Technology commented, "China's best scientists frequently collaborate with U.S. AI researchers, making it nearly impossible to exclude them from global AI developments."

In response, the U.S. government is working to finalize new regulations to restrict American investments in Chinese AI sectors potentially threatening national security. This development follows President Joe Biden’s October 2023 executive order aimed at balancing AI innovation with security safeguards.

As China advances its indigenous AI capabilities, questions about safeguarding intellectual property and restricting military applications of open-source models grow increasingly complex amid ongoing U.S.-China technology tensions.