US Tech Fears Chinese AI Model Distillation

According to Bloomberg, “adversarial distillation” has become a hot topic in Washington DC’s tech policy community. That is, attempts by foreign companies (specifically in China or with connections to the Chinese government) to quickly forge their own powerful, lower-cost variations of frontier AI models from US companies like OpenAI and Anthropic, by churning through millions of outputs from those models. The Claude makers have accused companies including Alibaba (makers of the Qwen suite of models) of carrying out an “industrial-scale attack” on their intellectual property, in violation of their terms of service. Apparently, lobbyists representing Big Tech have told lawmakers that large-scale distillation of US models by Chinese companies could pose an “existential threat” to the entire industry. They further warn that, once #Artificial Intelligence models can self-improve without human intervention, future distillation attacks may become impossible to monitor or prevent. But fear not, Americans: the Chinese embassy in Washington calls all of this a “groundless allegation”! So it’s probably fine. #🇨🇳 ChinA.I. 🤖🧠🦾🤖 https://www.bloomberg.com..