Washington, DC [US], September 9 (ANI): China-based artificial intelligence companies are carrying out aggressive, industrial-scale campaigns to extract capabilities from U.S. frontier AI models and use them to train their own models, according to a new Cybersecurity Advisory issued by the US National Security Agency (NSA), Federal Bureau of Investigation (FBI) and Cybersecurity and Infrastructure Security Agency (CISA).
The advisory, titled “China-Based Artificial Intelligence Companies Conducting Industrial-Scale Distillation Campaigns Against U.S. AI Companies,” said the activities are aimed at systematically extracting restricted proprietary functionalities and capabilities from US frontier AI models.
The agencies said AI knowledge distillation is itself a legitimate and useful technique in AI research.
However, they warned that China-based companies are using the technique against U.S. models in a way that allows Chinese AI models to continually close the technology gap without making the same level of research and development investments.
According to the advisory, such activities can reduce the need for large amounts of compute, electricity and foundational research that are required to develop frontier AI models from the ground up.
The U.S. agencies warned that advances in Chinese AI models could have wider national security implications, including by enhancing military and cyberattack capabilities that could potentially be used against the U.S. and its allies.
The advisory also highlighted the methods being used by China-based AI companies to avoid detection. These companies deliberately distribute their operations across the global AI ecosystem, including multiple AI model providers, cloud platforms and infrastructure providers.
The agencies said this approach helps the companies avoid creating a single point of detection while allowing them to systematically target the strongest capabilities and proprietary features of U.S. frontier AI models.
The U.S. agencies described the competition between the U.S. and China in frontier AI as having major national security implications. They said frontier AI models can pose significant risks across critical infrastructure, military, economic and technology domains.
The activity is also described as affecting the public sector, industry and foreign partner systems, as well as National Security Systems throughout the Defense Industrial Base and Department of War. This includes organisations that are actively using AI as well as those adopting AI software and tools, including large and small language models.
The advisory called for greater cooperation across the wider AI ecosystem to counter these activities. It said cloud providers, application programming interfaces aggregators and infrastructure providers can work together to identify and defend against industrial-scale AI knowledge distillation campaigns.
Cybersecurity analysts and other network defenders across the AI ecosystem have been advised to use the guidance issued by the agencies to identify related activity and implement recommended mitigations. (ANI)
US NSA and other agencies warn China AI firms copying capabilities from US models
