By James Pomfret and Jessie Pang

(Reuters) – Top Chinese research institutions linked to the People’s Liberation Army have used the publicly available Llama al Meta model to develop an AI tool for potential military applications, according to academic papers and analysts.

In a June paper reviewed by Reuters, six Chinese researchers from three institutions, including two from the People’s Liberation Army’s (PLA) main research body, the Academy of Military Sciences (AMS), detail how they used an early version of Meta’s Llama as a basis for what he calls “ChatBIT”.

The researchers used an earlier Llama 2 13B large language model (LLM), which Meta, incorporating its own parameters, to build a military-focused AI tool to collect and process information and provide accurate and reliable information for taking operational decisions.

ChatBIT has been fine-tuned and “optimized for dialogue and question-answering tasks in the military domain,” the paper said. It was found to outperform other AI models that were about 90% as capable as OpenAI’s powerful ChatGPT-4. The researchers did not detail how they defined the performance, nor did they specify whether the AI ​​model was put to work.

“This is the first time there is substantial evidence that China’s PLA military experts have systematically researched and attempted to harness the power of open-source LLMs, especially Meta’s, for military purposes,” said Sunny Cheung, associate fellow at The Jamestown Foundation, which specializes in China’s emerging and dual-use technologies, including AI.

The meta has embraced the open release of many of its AI models, including Llama. It imposes restrictions on their use, including a requirement that services with more than 700 million users apply for a license from the company.

Its terms also prohibit use of the designs for “military, war, nuclear industries or applications, espionage” and other activities subject to US defense export controls, as well as weapons development and content intended to “incite and promote violence” .

However, because Meta’s designs are public, the company has limited ways to enforce these provisions.

In response to questions from Reuters, Meta cited its acceptable use policy and said it took steps to prevent abuse.

“Any use of our designs by the People’s Liberation Army is unauthorized and against our acceptable use policy,” Molly Montgomery, Meta’s director of public policy, told Reuters in a phone interview.

Meta added that the United States needs to embrace open innovation.

“In the global AI competition, the supposed role of a unique and outdated version of an American open-source model is irrelevant when we know that China is already investing over a trillion dollars to overtake the US in AI,” Meta said. the spokesperson said in a statement.

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

“In the future, through technological refinement, ChatBIT will not only be applied to intelligence analysis, but also … strategic planning, simulation training and command decision-making will be explored,” the paper said.

China’s Defense Ministry did not respond to a request for comment, nor did any of the institutions or researchers.

Reuters could not confirm ChatBIT’s capabilities and computing power, although the researchers noted that its model incorporated only 100,000 military dialogue recordings, a relatively small number compared to other LLMs.

“It’s a drop in the ocean compared to most of these models (that) are trained with trillions of chips, so … it really makes me wonder what they’re actually doing here in terms of different capabilities,” said Joelle Pineau, vice president of the company. AI Research at Meta and Professor of Computer Science at McGill University in Canada.

The research comes amid a heated debate in US tech and national security circles about whether firms like Meta should make their designs publicly available.

US President Joe Biden signed an executive order in October 2023 that seeks to manage AI developments, noting that while there may be substantial benefits to the innovation, there are also “substantial security risks, such as removing safeguards within the model”.

This week, Washington said it was finalizing rules to curb US investment in artificial intelligence and other technology sectors in China that could threaten national security.

Pentagon spokesman John Supple said the Defense Department recognized that open-source models have both benefits and drawbacks, and that “we will continue to closely monitor and evaluate the capabilities of competitors.”

“COOKIE JAR”

Some observers say China’s progress in domestic AI development, including the establishment of dozens of research labs, has already made it difficult to prevent the country from closing the technology gap with the United States.

In a separate academic paper reviewed by Reuters, two researchers from the Aviation Industry Corporation of China (AVIC) – which the United States has designated a firm with ties to the PLA – described the use of the Llama 2 for “interference strategy training of airborne electronic warfare. “.

China’s use of Western-developed AI has also extended into domestic security. A paper in June described how Llama was being used for “intelligence policing” to process large amounts of data and improve police decision-making.

The state-run PLA Daily published a commentary in April on how AI could help “accelerate the research and development of weapons and equipment, contribute to the development of combat simulation and improve the efficiency of military training”.

“Can you keep them (China) out of the cookie jar? No, I don’t see how you can,” William Hannas, senior analyst at Georgetown University’s Center for Security and Emerging Technology (CSET), told Reuters. A 2023 paper by CSET found 370 Chinese institutions whose researchers have published work related to General Artificial Intelligence – helping to drive China’s national strategy to lead the world in AI by 2030.

“There is too much collaboration between China’s best scientists and the US’s best AI scientists for them to be left out of developments,” Hannas added.

(Additional reporting by Katie Paul in New York; Phil Stewart in Washington, Eduardo Baptista in Beijing and Greg Torode in Hong Kong; Editing by Gerry Doyle)