China's Liu Bin calls US AI distillation claims 'misguided'

Assistant Chinese Foreign Minister Liu Bin used the podium at the World AI Conference in Shanghai on Saturday to swat down a growing US narrative, telling attendees that "Some countries hype up distillation" and calling such claims "misguided and counterproductive," as Bloomberg reported. He did not name the United States directly. The intervention is Beijing's clearest response yet to accusations from Anthropic and others that Chinese labs are illicitly extracting results from top American models to train their own.

The backdrop is a specific, high-profile allegation. Anthropic has told the US Senate Banking Committee that Alibaba's Qwen lab ran roughly 28.8 million Claude queries through about 25,000 fake accounts earlier this year, an operation Anthropic frames as the largest known distillation campaign against its model. Distillation itself is not exotic tradecraft: it is the practice of using an older "teacher" model's outputs to train a cheaper "student" model that mimics its capabilities. What is contested is whether doing that against a competitor's paid API, at industrial scale and via fraudulent accounts, is unfair extraction or just aggressive competition. Alibaba denies wrongdoing.

The strategic signal from Shanghai matters more than the rhetoric. US frontier labs have started to coordinate on the problem, with reporting that OpenAI, Anthropic and Alphabet's Google are working together to clamp down on Chinese competitors allegedly harvesting their outputs. Xi Jinping, opening the same conference, pitched the opposite frame, saying AI development "should not be a solo performance by a single country." That is a policy fight, not a technical one, and it is now playing out simultaneously at WAIC, on Capitol Hill and inside the terms-of-service enforcement teams of the biggest model providers.

The honest caveat is that the underlying numbers are Anthropic's allegation, filed to a Senate committee, not an independent finding, and Bloomberg's report is careful to note Liu Bin did not name the US. What the reporting does not give you is the technical evidence tying that traffic specifically to Qwen, any concrete US regulatory response, or a signal on whether Beijing will issue guidance to its own labs. Watch instead for what US labs do next at the API layer, because tighter identity checks, rate limits and shared abuse signals are the levers that actually move before any policy does.