CISA, NSA and FBI Warn of China-Based AI Companies Targeting US AI Models with Industrial-Scale Knowledge Distillation Campaigns to Shortcut AI Development
U.S. cybersecurity agencies warn Chinese AI firms are using illicit knowledge distillation to extract proprietary U.S. AI models, accelerating their development at the expense of American innovation.
The Cybersecurity and Infrastructure Security Agency (CISA), National Security Agency (NSA), and Federal Bureau of Investigation (FBI) issued a joint advisory detailing China-based AI companies systematically targeting U.S. AI firms to extract proprietary models through knowledge distillation. This machine learning technique repurposes outputs from advanced models to train less capable ones, but is being exploited to shortcut legitimate AI development. The campaigns violate U.S. companies’ terms of service and undermine fair competition by enabling rapid advancement for Chinese firms.
The advisory identifies multiple China-based AI companies—including DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI—engaging in these activities since at least late 2024. These firms allegedly extracted billions of tokens across millions of exchanges from U.S. frontier AI models such as variants of Claude, GPT, Gemini, and Grok. The scale of these operations suggests coordinated efforts to accelerate AI capabilities within China’s AI development strategy.
CISA Acting Director Nick Andersen emphasized the urgency of safeguarding AI platforms against such distillation campaigns, which threaten to erode America’s competitive edge in AI innovation. The advisory calls on U.S. AI companies to implement stronger protections to prevent unauthorized extraction of proprietary models. The warning underscores the growing intersection of cybersecurity and AI development amid escalating global competition.
The joint advisory reflects broader concerns over China’s AI ambitions and the potential risks posed by state-aligned or state-supported entities. While knowledge distillation is a legitimate technique, its misuse in this context raises significant national security and economic competitiveness issues. The agencies urge vigilance and proactive measures to mitigate these threats to U.S. AI leadership.