商务部：蒸馏是AI领域各个模型间互相学习的通行做法，本质是中性技术手段

Odaily News: The spokesperson of the Ministry of Commerce answered reporters' questions regarding the U.S. announcement on cybersecurity related to Chinese AI enterprises' distillation activities in the U.S. The spokesperson stated that distillation is a common practice for AI models to learn from each other, and is essentially a neutral technical means, used by global model companies including U.S. firms. This technical approach helps models improve learning efficiency and achieve more efficient utilization of human knowledge, which has positive significance for countries developing their AI industries, unleashing the potential of AI technology, bridging the development gap, and more broadly benefiting developing countries and the general public.

The U.S. approach is a typical case of double standards. Regarding AI, China encourages open-source, open, and collaborative sharing. China's open-source models are open to global enterprises, including U.S. firms, facilitating their model research and development. U.S. companies' model R&D reports also disclose that they have extensively distilled Chinese models. In contrast, the U.S. has repeatedly and unilaterally accused Chinese enterprises of engaging in "industrial-scale" distillation, stigmatizing common industry practices as attacks. This reflects both the anxiety of the U.S. and its double standards.