U.S. security agencies, including the NSA, FBI, and CISA, stated on Tuesday that these companies, which also include DeepSeek and Alibaba, systematically extracted billions of tokens across millions of requests from models such as Claude, GPT, Gemini, and Grok since late 2024. They allegedly used proxy services and multiple accounts to avoid detection, a practice termed "malicious knowledge distillation." The agencies believe this occurred with the Chinese government's awareness. Moonshot's Kimi model had previously rattled markets in July by topping benchmarks against U.S. counterparts, leading to a selloff in Chinese AI stocks and broader semiconductor and crypto markets.