This week, the Chinese firm Moonshot AI unveiled the latest iteration of its Kimi model, igniting intense discussions surrounding China’s role in the open-source AI landscape.

Moonshot AI announced that while Kimi K3 “still lags behind leading proprietary models such as Claude Fable 5 and GPT 5.6 Sol,” the new open-source version “exhibited cutting-edge performance in our evaluation suite, consistently surpassing other models we tested.” Independent evaluations conducted by Arena.ai and Vals AI further indicated that Kimi is on par with prominent frontier AI models.

The release of this announcement coincided with Chinese President Xi Jinping’s address at the World AI Conference in Shanghai, a timing that seems to have rattled Wall Street; the Nasdaq fell approximately 1% on Friday as investors divested shares from semiconductor companies such as Nvidia.

Reactions from industry experts echo sentiments familiar to those who followed the discussions surrounding another Chinese entity, DeepSeek, which launched its open-source R1 model in January 2025. However, the current climate is amplified due to the ongoing trade tensions initiated during the Trump administration, ongoing debates regarding the national security implications posed by companies like Anthropic, and the impending IPOs from significant AI firms.

David Sacks, the former AI czar under the Trump administration and current co-chair of the President’s Council of Advisors on Science and Technology, compared Kimi’s advancements to the United States, which he described as “entangled in bureaucratic constraints: politicians and regulators are prohibiting new data centers, imposing state regulations, and advocating for new federal agencies to pre-approve frontier models. This is a recipe for losing the AI race.” He also seized the opportunity to criticize Anthropic, labeling Claude as an example of “woke, lobotomized models.”

Additionally, former Uber CEO Travis Kalanick echoed concerns regarding the notion that Chinese companies are “distilling off” (i.e., training on the outputs of) American AI models.

“If distillation isn’t regulated, then everyone should theoretically be able to extract knowledge from each other’s models. Otherwise, American models would be at a disadvantage,” Kalanick stated. (It is worth noting that American models have also drawn from Chinese ones, particularly Kimi.)

In the meantime, Dean Ball, OpenAI’s head of strategic futures, stated that Kimi is “a very impressive model” whose capabilities likely cannot be attributed solely to distillation or similar methods, expressing his surprise that the Chinese government continues to permit the open sourcing of such advanced models given the associated risks.

Ball further posited that “the likely outcome of a world dominated by open-weight models is a form of AI communism,” where AI becomes a “public good” provided by the state as a type of “digital public infrastructure.”

“This vision strikes me as a nightmarish reality, yet I have yet to meet an advocate for open-weight models who doesn’t ultimately recognize that this is where things are headed,” Ball remarked. He even suggested that the Trump administration (which he previously served) will eventually come to see the necessity of imposing substantial regulatory risks surrounding the use of open-weight Chinese models.

“There’s no need to ‘ban open source’ (a rather foolish theme in AI policy discussions),” Ball explained. “What’s required is to instruct every agency to implement soft laws that instill fear, uncertainty, and doubt. ‘A Federal Reserve Advisory Bulletin indicated that there may be vulnerabilities in Chinese AI models.’ This doesn’t even need to be thoroughly substantiated. The goal is to create enough regulatory apprehension that every regulated entity retreats.”

Conversely, Shakeel Hashim, editor of the AI-centric publication Transformer, contended that much of the anxiety surrounding Kimi is exaggerated, as the model “likely does not possess any dangerous cyber capabilities,” and noted that the Chinese government will face “very similar incentives” to limit open Chinese models once such capabilities are developed.