China's Stealth AI Threat: US Warns of 'Knowledge Distillation' Attacks

America's top security agencies — CISA, NSA, and FBI — issued a urgent warning on September 8, 2026, about a new and subtle threat from China: large-scale 'knowledge distillation' campaigns. Chinese AI companies are secretly copying the special abilities of advanced U.S. AI models, bypassing years of research and billions of dollars in development, likely with their government's support. This isn't about killer robots, but a quieter battle for control over how technology shapes our world. Chinese firms like DeepSeek and Moonshot AI have taken billions of 'tokens' of information from leading U.S. models like GPT, Claude, Gemini, and Grok since late 2024. They do this by using clever tricks like hiding their activities through many different computer pathways, making it hard to stop them. Meanwhile, China's spy chief, Chen Yixin, has also warned that U.S. AI models could threaten the Communist Party's power through 'cognitive warfare,' which means using AI to spread misleading information and control how people think. The U.S. is fighting back, with CISA asking American AI companies to find and stop these copying attempts, even by slightly changing how their AI models respond to suspicious questions. U.S. Treasury Secretary Scott Bessent also warned that Chinese companies involved in this 'IP theft' could face serious penalties, like economic restrictions. These moves show how serious the competition is, as both countries try to lead in the fast-changing world of Artificial Intelligence, with the U.S. even suggesting a special 'hotline' to talk about dangerous AI incidents.