Chinese military used OpenAI, Anthropic AI outputs to train their defence systems: Report
Chinese military researchers used outputs from OpenAI and Anthropic AI models to train domestic AI systems for defence, according to a Reuters review of over 80 Chinese academic papers and patents.
The technique, model distillation, uses outputs from powerful AI models to train smaller, specialized models that can run locally with fewer computing resources.
Researchers from PLA Unit 96941, a Beijing-based military intelligence and cyber warfare unit, used OpenAI's GPT-3.5 to process sensitive military software code, per a paper published last year.
The findings were based on research compiled by the Washington-based Jamestown Foundation and included analysis of more than 60 papers by fellow Sunny Cheung.
US officials have accused some Chinese organizations of using distillation to extract capabilities from American AI systems, while China has denied the allegations.
Story angle
Center-right
Mostly straight facts, with a light government- or market-friendly tilt.
About how the story is framed, not whether the facts are true.
Times of India
Center-right
Mixed factuality
Who owns this?
Times Group
Bennett, Coleman & Co. (Sahu Jain family)
India's largest newspaper group by circulation.
Wikipedia: Sahu Jain family