Chinese AI firms are siphoning capabilities from American models, CISA warns – Help Net Security

Chinese AI firms are siphoning capabilities from American models, CISA warns – Help Net Security
U.S. agencies say China-based AI companies have been running large-scale knowledge distillation campaigns to copy advanced capabilities from frontier models such as Claude, GPT, Gemini, and Grok. The advisory names DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI, and warns providers to strengthen verification, monitoring, and output controls to slow extraction efforts. #DeepSeek #MoonshotAI #AlibabaGroup #MiniMax #StepFun #ZAI #Claude #GPT #Gemini #Grok #CISA #NSA #FBI

Keypoints

  • China-based AI firms are allegedly using industrial-scale knowledge distillation to copy U.S. frontier model capabilities.
  • DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI are named as major actors in the advisory.
  • The campaigns have extracted billions of tokens through millions of requests since at least late 2024.
  • Transfer stations, premium account pools, and traffic routing are used to bypass restrictions and hide origins.
  • Agencies recommend stronger identity checks, monitoring, response variation, and differential privacy to reduce extraction.

Read More: https://www.helpnetsecurity.com/2026/09/09/china-malicious-ai-knowledge-distillation-against-us-companies/