Moonshot's Kimi K3 stokes new US fears over Chinese open-source AI

Chinese company Moonshot AI released a new version of its Kimi model this week, Kimi K3, prompting a fresh round of debate about China’s progress in open-source AI. Moonshot said Kimi K3 still trails the most powerful proprietary models, Claude Fable 5 and GPT 5.6 Sol, but that it has been “consistently outperforming other tested models” on the company’s own evaluation suite. Independent analyses from Arena.ai and Vals AI also suggested Kimi is competitive with flagship frontier models. The release coincided with a speech from Chinese president Xi Jinping at the World AI Conference in Shanghai, and appeared to unsettle Wall Street: the Nasdaq dropped about 1% on Friday as investors sold off stocks in chip companies including Nvidia.

The reaction echoes the debate that followed DeepSeek’s open-source R1 model release in January 2025, though tech-industry figures now frame it against the Trump administration’s tariff war with China and repeated fights over the national security threat supposedly posed by Anthropic. David Sacks, the administration’s former AI czar and now co-chair of the President’s Council of Advisors on Science and Technology, argued that U.S. politicians and regulators are creating unnecessary obstacles — banning new data centers, layering on state regulations, and pushing for federal pre-approval of frontier models — and said that approach risks losing the AI race to China.

OpenAI’s head of strategic futures, Dean Ball, said Kimi is a strong model whose performance likely isn’t explained away by distillation from American models. He warned that a world dominated by open-weight models risks drifting toward what he called an AI-communism model, where the state ultimately treats AI as public infrastructure, and predicted the Trump administration will eventually generate regulatory uncertainty around open-weight Chinese models without formally banning them. Shakeel Hashim, editor of the AI-focused publication Transformer, pushed back on the alarm, arguing Kimi is unlikely to have dangerous cyber capabilities and that China’s government will face similar incentives to restrict open models once they do.

Why it matters

The split among senior industry and policy figures — over whether an increasingly capable Chinese open-source model is an economic challenge, a national-security threat, or neither — suggests how Washington treats open-weight AI could become a live regulatory battleground rather than a settled question.