Meta’s Llama used by Chinese researchers to build AI model for military use – report

Meta’s publicly available AI model Llama model was used by some Chinese research institutes associated with the People’s Liberation Army to develop an AI product for potential military uses, Reuters reported, citing academic papers and analysts.

Six Chinese researchers from three institutions, including two under the PLA’s research organization known as Academy of Military Science, or AMS, showed how they used an early version of the large language model, or LLM, Llama as a base for what it calls ChatBIT, the report added citing June research paper.

Llama 2 13B was used by the researchers to include their own guidelines to develop a military-focused AI tool to collect and process intelligence, and provide accurate and dependable information for operational decision-making, the report noted.

ChatBIT was customized and “optimized for dialogue and question-answering tasks in the military field,” according to the paper. The product was seen to outperform some other AI models which were about 90% as capable as OpenAI’s ChatGPT-4.

The researchers did not disclose how they measured performance or specify if the AI model had been put into service, the report noted.

Meta does have restrictions on the use of its LLMs which included a requirement that services with over 700 million users avail a license from the company.

The U.S. social media giant also forbids the use of the models for “military, warfare, nuclear industries or applications, espionage, use for materials or activities that are subject to the International Traffic Arms Regulations maintained by the United States Department of State.”

“Any use of our models by the People’s Liberation Army is unauthorized and contrary to our acceptable use policy,” said Molly Montgomery, Meta’s director of public policy, in an interview to the news agency.

The Chinese researchers include Geng Guotong and Li Weiwei with the AMS’s Military Science Information Research Center and the National Innovation Institute of Defense Technology, and researchers from the Beijing Institute of Technology and Minzu University.

ChatBIT’s capabilities and computing power could not be known, but researchers noted that its model includes only 100,000 military dialogue records, a small number versus other LLMs, the report added.

“That’s a drop in the ocean compared to most of these models [which] are trained with trillions of tokens so … it really makes me question what do they actually achieve here in terms of different capabilities,” said Joelle Pineau, a vice president of AI Research at Meta and a professor of computer science at McGill University in Canada, according to the report.

In a separate academic paper, two researchers with the Aviation Industry Corporation of China — which the U.S.has designated as having links to the PLA — described using Llama 2 for “the training of airborne electronic warfare interference strategies,” the report stated.

In addition, a June paper showed how Llama had been used for “intelligence policing” to process large volumes of data and boost decision-making.

Generative AI services have become the talk of the town since the launch of Microsoft (MSFT)-backed OpenAI’s ChatGPT in 2022. Globally, companies have launched their own LLMs, which can provide services such as content, image, video and voice generation, to name a few.

Chinese companies are not far behind in developing their LLM powered AI chatbots, such as Alibaba’s (BABA) Qwen2.5, Tongyi Qianwen 2.0, and Tongyi Wanxiang and Baidu’s (BIDU) Ernie Bot.

In May it was reported that a LLM was being trained on President Xi Jinping’s political philosophy known as “Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era,” and other official literature provided by the Cyberspace Administration of China.