Kimi: A threat or a threat? | TechCrunch

Chinese company Moonshot AI released a new version of its Kimi model this week, prompting a wave of perhaps inevitable rhetoric about China and open source AI.

Moonshot said that although the Kimi K3 “still follows the strongest proprietary models, Claude Fable 5 and GPT 5.6 Sol,” the new open-source model “demonstrated border-level performance across our evaluation suite, consistently outperforming the other models tested.” Independent analyzes from Arena.ai and Vals AI also indicated that Kimi is competitive with leading frontier models.

The announcement, which coincided with a speech by Chinese President Xi Jinping at the World Conference on Artificial Intelligence in Shanghai, appeared to spook Wall Street, with the Nasdaq index falling about 1% on Friday as investors sold shares in chip companies such as Nvidia.

Many of the resulting posts from tech industry figures will sound familiar to those who remember the debate after another Chinese company, DeepSeek, released an open-source R1 model in January 2025. Except now, everything seems high after the Trump administration’s tariff war with China, repeated battles over the national security threat Anthropic supposedly poses, and with major AI companies preparing to finally go public.

For example, David Sachs, Trump’s former AI czar and now co-chair of the President’s Council of Advisors on Science and Technology, compared Kimi’s progress to a United States that “is holding itself back: Politicians and bureaucrats block new data centers, impose state-level regulations, and pressure new federal agencies to pre-approve frontier models. That’s how you lose the AI ​​race.” (The news also gave him an excuse to dig into Anthropics, calling Claude an example of “woke archetypes” who represent “the enemy of American competitiveness.”)

Travis Kalanick, former CEO of Uber, echoed complaints that the Chinese are “distilling” (i.e., training on the outputs) of American AI models.

“If distillation is not enforced against everyone, then everyone should be able to distill from everyone…otherwise one arm [would be] “They are tied behind the backs of the American models,” Kalanick wrote. “(Of course, the American models were also built on top of the Chinese models, specifically Kimi).”

Meanwhile, Dean Paul, head of strategic futures at OpenAI, said Kimi is a “very good model” and its performance probably “cannot be explained by distillation or anything like that,” adding that he is “personally surprised that the Chinese state continues to allow open sourcing of models of this quality, given the potential risks.”

In fact, Paul has suggested that “the likely outcome of a world dominated by the open weight model is full-blown AI communism,” where AI is treated as a “public good” that will eventually be provided by the state as a kind of “digital public infrastructure.”

“This future looks to me like a dystopian hellscape, but I’ve never met an open-weight model advocate who doesn’t ultimately acknowledge that this is where things end,” Paul said. He even suggested that the Trump administration (for which he worked) would eventually realize that it needed to “create significant amounts of regulatory risk around the use of Chinese open-weight models.”

“You don’t need to ban open source (one of the dumber ideas in the AI ​​policy debate),” Paul said. “You just need to direct every agency to issue a soft law that leads to FUD [fear, uncertainty, and doubt]. “A Federal Reserve advisory found there may be backdoors in Chinese AI models.” It doesn’t have to be well justified. You create just enough regulatory risk that every regulated organization backs off.

However, Shakeel Hashim, editor of the AI-focused Transformer magazine, says many of the fears are overblown, because Kimi is “unlikely to possess serious cyber capabilities,” and because the Chinese government would face “very similar incentives” to restrict open Chinese models once those capabilities are developed.