Chinese military adapts Meta’s Llama for AI tool

China’s People’s Liberation Army (PLA) has adapted Meta’s open-source AI model, Llama, to create a military-focused tool named ChatBIT. Developed by researchers from PLA-linked institutions, including the Academy of Military Science, ChatBIT leverages an earlier version of Llama, fine-tuned for military decision-making and intelligence processing tasks. The tool reportedly performs better than some alternative AI models, though it falls short of OpenAI’s ChatGPT-4.

Meta, which supports open innovation, has restrictions against military uses of its models. However, the open-source nature of Llama limits Meta’s ability to prevent unauthorised adaptations, such as ChatBIT. In response, Meta affirmed its commitment to ethical AI use and noted the need for US innovation to stay competitive as China intensifies its AI research investments.

China’s approach reflects a broader trend, as its institutions reportedly employ Western AI technologies for areas like airborne warfare and domestic security. With increasing US scrutiny over the national security implications of open-source AI, the Biden administration has moved to regulate AI’s development, balancing its potential benefits with growing risks of misuse.