WASHINGTON, September 8, 2026 — The U.S. National Security Agency, Federal Bureau of Investigation and Cybersecurity and Infrastructure Security Agency released a joint cybersecurity advisory accusing several mainland Chinese artificial-intelligence companies of conducting aggressive, industrial-scale model-distillation campaigns against U.S. frontier AI companies.
The agencies say the activity is intended to extract restricted proprietary capabilities from American models and use those outputs to accelerate the training of competing systems. Reuters reported that the companies named include DeepSeek, Moonshot AI, Alibaba, MiniMax and StepFun.
Distillation itself is a widely used and legitimate machine-learning technique. The U.S. agencies’ allegation is narrower: they contend that the cited firms used distillation in a targeted way to obtain restricted proprietary functionality and capabilities from U.S. frontier models. Those allegations should be treated as official U.S. government findings and claims, not as court-adjudicated facts.
The NSA said such activity can help mainland Chinese AI developers narrow the technology gap while avoiding some of the research, computing and electricity costs required to build frontier systems from scratch. The advisory also argues that advances in those systems can affect military, cyber and critical-infrastructure capabilities.
Reuters reported that U.S. officials said the activity likely occurred with awareness of the Mainland Chinese Communist Government. That is an attributed U.S. assessment; it is not an independently established fact. The report also said the Chinese embassy in Washington had not responded to the allegations at the time of publication.
Related coverage
Editorial: AI distillation is not the issue — covert extraction of restricted capabilities is
Rapid Response: U.S. agencies warn of industrial-scale AI model distillation