Chinese search giant Baidu denied any affiliation with a military research laboratory which accessed its publicly available generative AI model along with others to teach an experimental system.

South China Morning Post reported researchers in the country are using large language models (LLMs) including Baidu’s Ernie Bot and iFlyTek’s Spark to train a military AI platform about facing unpredictable human enemies.

In a statement, Baidu insisted it had “not engaged in any business collaboration or provided any tailored service” to the researchers or “any institutions with which they are affiliated”.

The company noted Ernie Bot is available to the general public.

In a research paper published by scholars at a Chinese university, they describe how they built prompts and received responses from LLMs using the functions available to any user interacting with generative AI tools. 

Richard Windsor, founder of the blog Radio Free Mobile, described the link between Ernie Bot and the People’s Liberation Army (PLA) as “tenuous at best”.

Windsor added the idea Baidu “is in cahoots with the Chinese military as a result of this is absurd”, noting the owners of publicly-available LLMs have no control over who uses them.

The PLA’s Strategic Support Force used Ernie Bot to examine possibilities of using AI in certain military scenarios. Windsor believes the researchers also accessed LLMs from OpenAI, Meta Platforms, Google and Anthropic.

Windsor lamented China’s technology sector is being targeted externally by the US, limiting access to advanced technology, while internally the Chinese government is “dissuading foreign capital from investing”.