US Agencies Accuse China-Based AI Companies of Distillation
The NSA, CISA, and FBI announced that Chinese AI companies are conducting large-scale unauthorized data scraping campaigns targeting American models.
The US National Security Agency, CISA, and the FBI issued a joint statement alleging that certain China-based artificial intelligence companies are conducting industrial-scale distillation campaigns against American frontier models and systems. It was emphasized that these activities pose national security risks.
Statement by US Agencies
In a joint statement released by the National Security Agency, CISA, and the Federal Bureau of Investigation, the activities of China-based firms were strongly criticized. Officials stated that these initiatives are targeted and aggressive in nature.
Among the allegations were that these processes likely took place with the knowledge of the Chinese government and targeted the American artificial intelligence ecosystem.
Targeted Companies and Models
It was claimed that companies mentioned in the statement, such as DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI, have scraped billions of tokens since late 2024.
It was noted that leading US AI models, including Claude, GPT, Gemini, and Grok, were targeted via millions of requests.
Distillation Technology Discussions
Distillation, widely used in artificial intelligence development, enables the capabilities of a more powerful teacher model to be transferred to a smaller student model.
However, leading laboratories such as Anthropic and OpenAI have long been trying to implement various security measures to prevent the unauthorized use of this method.
Future Discussions and Reflections
It drew attention that these developments coincided just ahead of planned meetings between US Treasury Secretary Scott Bessent and Chinese officials to address artificial intelligence security issues.
There are differing approaches among industry executives regarding the scale of concerns surrounding distillation.