Meta Llama 13B repurposed by China's PLA

China’s People’s Liberation Army (PLA) has modified Meta’s open-source Llama 13B model, even though Meta has prohibited military use. Artificial Intelligence (AI) known as “ChatBIT,” is being used for strategic planning, gathering intelligence, training and decision-making. 

PLA-linked Chinese researchers published details about the ChatBIT, an AI model that trains 100,000 military dialogue records and is highly capable of processing defence data. Chinese aviation firm associated with the PLA is another Llama-based model used for electronic warfare training. 

Meta CEO Mark Zuckerberg argues that restriction of open-source AI would harm the US, rather than benefit it. He recommends that open-source AI is beneficial in promoting transparency and a path to global advancement. 

Policy violations and risks

Meta’s guidelines do not allow the use of AI as in the military, but the open-source nature makes this policy challenging.

National security concerns

Meta’s representative cautioned that restricting open-source AI can jeopardise American leadership in AI, which will make the whole country’s economy and security bear consequences. Moreover, Meta has control over the third-party apps because the model already exists in public.

Open-source AI governance poses challenges around balancing innovation and the security threats that come with its advancement.