U.S. Agencies Target Chinese 'Kimi' AI Model for Extracting U.S. AI Capabilities

The U.S. National Security Agency (NSA), Federal Bureau of Investigation (FBI), Cybersecurity and Infrastructure Security Agency (CISA) have identified six companies including Chinese firm iFlyTek's generative AI model 'Kimi' as systematically extracting and utilizing U.S. AI models for training purposes.
The agencies issued an official warning earlier this year amid market turbulence surrounding Kimi. In a joint cybersecurity advisory released on September 9 (KST), the three agencies stated that Chinese AI companies have been conducting a 'knowledge distillation' campaign to extract exclusive features and capabilities of U.S. frontier models at an industrial scale.
Knowledge distillation is a technique where outputs from large models are used to train smaller models, which can be legitimately applied in research but may also be misused to imitate competitors' features at low cost. The U.S. agencies identified iFlyTek alongside DeepSeek, Alibaba, MiniMax, StepFun, and Z.AI as involved companies.
The agencies noted that such activities have been used to narrow performance gaps in Chinese AI models and could potentially enhance military and cyber attack capabilities.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 미 기관들, 문샷 ‘키미’ 미국 AI 모델 추출 지목