Three US agencies have named DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI as allegedly distilling American AI models at industrial scale since late 2024. The agencies also disputed DeepSeek's claim that its model was trained for about $5.6 million. They advised US providers to quietly reduce the quality of responses served to accounts flagged as suspicious.
A discussion thread highlights that Alibaba's Qwen 3.8 appears to reproduce the reasoning-prefill behaviour seen in GPT-5.5 Pro. Reasoning prefills let a user seed part of the model's chain of thought to steer how it solves a task. The report gives no benchmark numbers or confirmation from the vendors, so the comparison remains anecdotal.
US authorities say Chinese AI developers including DeepSeek and Alibaba are systematically pulling knowledge out of American models such as those built by OpenAI and Google. The agencies frame the activity as industrial espionage aimed at closing the capability gap at low cost. No details were given on which specific countermeasures might follow.