Kimi K3: The 2.8-Trillion-Parameter Open-Weight AI Model That Sparked a US-China IP Showdown

The U.S. Accusations and Sanctions Threat

On July 22, the White House struck. Michael Kratsios, the White House Science and Technology Adviser, publicly accused Moonshot of conducting "large-scale, covert industrial distillation" of Anthropic's most advanced model, Claude Fable 5, to build K3 . Kratsios alleged that Moonshot built a "sophisticated internal platform" to systematically extract capabilities from Fable via distillation — a common training technique where a smaller model learns from the outputs of a larger model .

The accusation went further. Kratsios also alleged Moonshot acquired restricted Nvidia GB300 chips via servers in Thailand to train K3, violating U.S. export controls . Treasury Secretary Scott Bessent doubled down, stating on Fox Business that sanctions "remain on the table" for Chinese AI companies that steal American IP, explicitly referencing the Moonshot case . "Open source is not open season on American IP," Bessent posted on X .

Denials and Pushback

Moonshot AI denied the allegations, stating it developed K3 through independent research and innovation . China's government formally rejected the U.S. claims, accusing Washington of making "politically motivated accusations" and emphasizing that distillation is a standard, widely used AI training technique that does not constitute IP theft under current law .

Many AI experts also poured cold water on the White House's claims. Skeptics noted that Anthropic's Fable model was only publicly available for about two weeks before Kimi K3's launch — an implausibly short window for the scale of distillation Kratsios described .

Industry Pushback: 25 Companies Sign Open Letter

On July 24-25, 2026, a coalition of 25 companies — including Nvidia, Microsoft, Meta, IBM, Palantir, Hugging Face, Mistral, and Perplexity — signed a joint open letter titled "Open Weights and American AI Leadership" . The letter urged U.S. policymakers not to impose broad, premature restrictions on open-weight AI models. It defended distillation as "a decades-old development technique" and argued that restricting open models would harm U.S. competitiveness and innovation .

Notable details from the coalition:

• OpenAI quietly signed the letter, though its closed models would not be directly affected .
• Anthropic and Google did not sign .
• Nvidia CEO Jensen Huang marked his first-ever post on X by sharing the letter, stating that "open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and empower small companies, researchers, and startups" .

The pushback came in direct response to the White House's distillation accusations and broader calls for new export controls on open model weights .

A "DeepSeek Moment" for Open-Source AI?

Multiple outlets and analysts have described K3's release as a potential "DeepSeek Moment" for open-source AI, drawing a direct parallel to the shockwave that DeepSeek's R1 model sent through the industry earlier in 2026 . Like R1, K3 demonstrated that a Chinese open-weight model could match or approach the performance of top U.S. proprietary systems, while being freely available — disrupting assumptions about the necessity of closed, frontier models and triggering both market panic and geopolitical friction .

The term "DeepSeek Moment" describes a sudden recognition that open Chinese AI has reached parity with the U.S. frontier, and K3's scale and timing have been widely framed in that context.

What Happens Next?

The full model weights of Kimi K3 were released on July 27, 2026, as promised, making the model freely downloadable for developers and researchers . The U.S. government has not yet taken formal enforcement action, but the threat of sanctions and Entity List designations remains active. The incident has become a flashpoint in the broader U.S.-China AI rivalry, with open-weight models at the center of the debate over innovation, national security, and intellectual property.

For now, Kimi K3 stands as the largest open-weight AI model ever created — and as a symbol of the high-stakes tension between openness and control in the age of frontier AI.