Kimi K3 reignites the China AI panic

Moonshot AI’s release of Kimi K3 has reopened a familiar argument about China, open-source models, and the United States' ability to compete in frontier AI.

Moonshot said Kimi K3 “still trails the most powerful proprietary models, Claude Fable 5 and GPT 5.6 Sol,” but claimed the model showed “frontier-level performance across our evaluation suite, consistently outperforming other tested models.” Independent analyses from Arena.ai and Vals AI also found Kimi competitive with flagship frontier models.

The announcement coincided with a speech by Chinese President Xi Jinping at the World AI Conference in Shanghai. It appeared to unsettle Wall Street: the Nasdaq fell about 1% on Friday, while investors sold shares in chip companies including Nvidia.

Recommended reading

Nubia’s iMoochi robot pet has OLED eyes and 4G

Kimi K3 revives the DeepSeek debate

The reaction echoes the market and policy debate that followed DeepSeek’s release of its open-source R1 model in January 2025. This time, however, the stakes are heightened by the Trump administration’s tariff war with China, repeated disputes over the national security threat allegedly posed by Anthropic, and preparations by major AI companies to go public.

David Sacks, the Trump administration’s former AI czar and current co-chair of the President’s Council of Advisors on Science and Technology, contrasted Kimi’s progress with what he described as a United States “tying itself in knots.”

“Politicians and bureaucrats are banning new data centers, piling on state regulations, and pushing for new federal agencies to pre-approve frontier models. This is how you lose the AI race.”

Sacks also used the news to criticize Anthropic, describing Claude as an example of “woke lobotomized models.”

Former Uber CEO Travis Kalanick focused on model distillation—the practice of training a system on the outputs of another model. He argued that Chinese developers are “distilling off” American AI systems, while acknowledging that American models have also been built on Chinese ones, specifically Kimi.

“If distillation isn’t enforced against, then everyone should be able to distill from everyone else.. otherwise one arm [would be] tied behind American models' backs.”

Open weights and regulatory pressure

OpenAI head of strategic futures Dean Ball called Kimi “a very good model” and said its performance probably could not be “explained away by distillation or anything like that.” He added that he was “personally surprised the Chinese state continues to allow the open sourcing of models this good, given potential risks.”

Ball predicted that an open-weight-model-dominant future could become “full AI communism,” with AI treated as a public good and eventually supplied by the state as “digital public infrastructure.” He described that prospect as a “dystopian hellscape” and argued that the Trump administration could eventually create substantial regulatory risk around Chinese open-weight models.

His proposed approach would not directly ban open source. Instead, agencies could issue guidance designed to create fear, uncertainty, and doubt—for example, suggesting that Chinese models might contain backdoors—until regulated companies avoid them.

Shakeel Hashim, editor of the AI publication Transformer, said the concern is overstated. He argued that Kimi “likely does not have dangerous cyber capabilities” and that the Chinese government would face “extremely similar incentives” to restrict open Chinese models if they eventually gained those capabilities.