US administration accuses six Chinese AI companies of “distillation”, escalating US-China conflict

[Washington Current Affairs] On the 8th, the Trump administration released a statement criticizing six companies, including Chinese startup Moonshot AI, for ``distilling'' models developed by American artificial intelligence (AI) companies. The US administration has become increasingly wary of China's rapid development of the latest AI models, and tensions have intensified ahead of the US-China summit expected on the 24th.

Huawei to adopt AI semiconductors, China deep seek, away from dependence on US

Distillation is a method that allows AI to learn the output of large-scale models, and can significantly reduce development time and costs.

According to the statement, startup DeepSeek, tech giant Alibaba Group, and Beijing Zhifuhuazhang Technology, also known as Z.ai, will send millions of requests to advanced U.S. models starting in the second half of 2024. The company claims to have collected billions of responses regarding tokens, which are units used by AI to understand and generate sentences.

Models such as American Anthropic's ``Claude'', Open AI's ``Chat GPT'', Alphabet Inc.'s Google ``Gemini'', and SpaceX's ``Grok'' were used.

"China-based AI companies are conducting an aggressive and malicious distillation on an industrial scale, distilling the limited and unique capabilities and capabilities of America's most advanced AI models," the statement said. ``It is highly likely that this was done with the knowledge of the Chinese government.''