Chinese open-source AI model Kimi K3 sparks debate and market jitters
Moonshot AI's new Kimi K3 model, competitive with top proprietary models, has caused a Nasdaq dip and reignited debates over AI regulation, national security, and the future of open-source AI.
Chinese company Moonshot AI released a new version of its Kimi K3 model this week, an open-source model that the company says, while still trailing the most powerful proprietary models such as Claude Fable 5 and GPT 5.6 Sol, demonstrates frontier-level performance and consistently outperforms other tested models. Independent analyses from Arena.ai and Vals AI also indicate that Kimi is competitive with leading frontier models.
The announcement coincided with a speech by Chinese President Xi Jinping at the World AI Conference in Shanghai. The news appears to have spooked Wall Street, with the Nasdaq dropping about 1% on Friday as investors sold off stocks in chip companies like Nvidia. Many comments from tech industry figures echo debates that followed another Chinese company's DeepSeek R1 release in January 2025. The situation is now heightened amid the Trump administration's tariff war with China, repeated fights over the national security threat posed by Anthropic, and as major AI companies prepare to go public.
David Sacks, the Trump administration's former AI czar, criticized US regulation, saying the country is "tying itself in knots" by banning new data centers and imposing state regulations. He also took a dig at Anthropic, calling Claude an example of "woke lobotomized models." Former Uber CEO Travis Kalanick accused the Chinese of "distilling" from American AI models, but noted that American models have also been built on top of Chinese ones, specifically Kimi.
Dean Ball, OpenAI's head of strategic futures, said Kimi is "a very good model" whose performance probably cannot be explained away by distillation. He suggested that the probable outcome of an open-weight-model-dominant world is "full AI communism," where AI is treated as a state-provided "digital public infrastructure." Ball proposed that the Trump administration should create regulatory risk around Chinese models, for example by issuing soft law that creates fear, uncertainty, and doubt about backdoors. However, Shakeel Hashim, editor of the AI publication Transformer, argued that such worries are overblown, both because Kimi likely does not have dangerous cyber capabilities, and because the Chinese government will face similar incentives to restrict open models once they develop those capabilities.
