Chinese researchers linked to the People's Liberation Army (PLA) have developed an AI model called ChatBIT, designed for military applications using the open-source Meta Llama model. According to ReutersSome of the researchers are associated with the Academy of Military Sciences (AMS), the PLA's main research group.

Three academic papers and several analysts have confirmed this information, highlighting that ChatBIT uses Meta’s Llama 13B large language model (LLM). This LLM has been modified for intelligence gathering and processing, allowing military planners to use it for operational decision making.

According to one of the articles cited by Reuters, the military AI is “optimized for military dialogue and question-answering tasks.” ChatBIT was also claimed to perform around 90% of the performance of OpenAI’s LLM GPT-4, though the article did not indicate how its performance was tested or whether the AI model has been used in the field. However, its use of open-source AI models could allow it to match the latest models released by US tech giants in benchmark tests.

“This is the first time there is substantial evidence that PLA military experts in China have been systematically researching and attempting to harness the power of open-source LLMs, especially Meta’s, for military purposes,” says Sunny Cheung, a research associate at the Jamestown Foundation, a Washington DC-based think tank that focuses on China’s emerging and dual-use technologies, including artificial intelligence. Meta’s license explicitly prohibits the use of Llama for military applications, but its open-source nature makes it nearly impossible to enforce such limits in reality.

However, Meta said in a statement that this alleged use of the LLM Llama 13B, which it considers an “outdated version” given that it is already training Llama 4, is largely irrelevant, especially given that China is investing billions of dollars to gain an edge in AI technologies. Furthermore, other researchers pointed out that ChatBIT only used 100,000 military dialogue records, a minimum considering that the latest models are trained with trillions of data points.

Some experts question the viability of such a small dataset for military AI training. However, ChatBIT could also just be a proof of concept, with the military research institutes involved planning to create more extensive models. Furthermore, the Chinese government could have released these research papers as a signal to the US that it is not afraid to use AI to gain a technological advantage on the global stage.

No matter how big or small this development is, Washington has been wary of this news: the use of American open-source technologies that will give a military advantage to its opponents. Therefore, in addition to expanding export controls on China, many US lawmakers also want to block the country's access to open-source/standardized technologies such as RISC-V. It is also taking steps to prevent US entities from investing in AI, semiconductors, and quantum computing in China.

This is the double-edged sword that American policymakers must contend with. Naturally, they do not want to give access to advanced technologies to their opponents through open source avenues; however, open source technology is also a major driver of technological advances, and limiting it could put American companies at a disadvantage.