Recent years have seen a shift in US policy to block China's access to advanced chips and artificial intelligence technology. The government hoped it could prevent China from using this cutting-edge tech to build military systems, but the cat is out of the bag. A new report claims top Chinese research facilities with ties to the country's government have developed a version of Meta's Llama AI for military use.

This report comes from Reuters, which claims to have reviewed research papers from June of this year. The documents detail work conducted by six Chinese computer scientists, including Geng Guotong and Li Weiwei from the army's Academy of Military Science. The team reportedly used the Llama 13B model as a base upon which they built an AI that can collect, process, and analyze military intelligence.

The military AI, known as ChatBIT, is essentially a chatbot. However, general tools like ChatGPT lack the specificity of ChatBIT, which was optimized to answer questions and provide guidance for military decisions. The guardrails that might prevent the average chatbot from recommending violence or unethical behavior have been removed, and Chinese authorities don't seem bothered that ChatBIT blatantly violates Meta's Llama terms.

Meta makes multiple AI models available freely, including Llama 13B. The name denotes the size of the model -- 13 billion parameters in this case. This isn't the largest version of Llama, which goes up to 65 billion parameters. Meta prohibits Llama from being used for "military, warfare, nuclear industries or applications, espionage" and anything else that could run afoul of US export controls. However, Meta has no way of preventing this kind of use with publicly available models.

While Llama 13B is capable of ingesting a large volume of data, the study apparently only used 100,000 military records. This suggests the model is in the earliest stages, but the team hopes to train it to assist with "strategic planning and command decision-making," reports Reuters.

Also unclear is the amount of processing power behind ChatBIT. Washington has gone to great lengths to limit the availability of AI accelerators in China, leading domestic manufacturers to move aggressively to offer scalable alternatives. Chinese authorities are not bothering to keep ChatBIT a secret, but no doubt more advanced military AI projects are still under wraps. Even with a shortage of AI hardware, the Chinese government can undoubtedly pull together enough power to work with large AI models in a military context.

Despite the US government's efforts, the existence of ChatBIT suggests Washington won't be able to slow China's AI progress for much longer. China has set a goal of leading the world in AI by 2030, and it's not letting American roadblocks get in the way.