The Trump administration is weighing entity list restrictions against Chinese AI labs as a 2.8-trillion-parameter model threatens American dominance
A White House official has accused Moonshot AI, the Beijing-based startup behind the Kimi K3 model, of improperly using American AI models and Nvidia chips to build what is now the largest open-weight AI system on the planet.
Kimi K3 launched on July 17 with 2.8 trillion parameters and promptly started matching or closely rivaling systems from Anthropic and OpenAI on key benchmarks, particularly in coding tasks. Demand was so intense that Moonshot had to halt new subscriptions almost immediately after launch.
What the White House is actually alleging #
The core accusation is twofold. First, that Moonshot used Nvidia chips that should have been restricted under existing export controls. Second, that the company employed AI distillation techniques on US-built models, essentially learning from American systems to train its own.
David Sacks, an AI advisor to the Trump administration, called the situation an “inflection point” in AI policy regarding competition with open-source technologies.
Officials from US labs, including Anthropic, have separately raised concerns about potential intellectual property theft and the circumvention of chip export controls.
The Trump administration is reportedly considering placing Chinese AI labs on the Commerce Department’s Entity List.
Why this matters for markets #
Nasdaq indices dipped following the Kimi K3 launch, reflecting investor anxiety about US firms potentially losing their competitive moat.
The bigger picture #
What makes Kimi K3 different is scale. At 2.8 trillion parameters, it is the largest open-weight model in the world. The fact that it’s open-weight, meaning its architecture is publicly accessible, adds another layer of complexity. Open-weight models are harder to contain through traditional export controls because the weights themselves can be shared freely once released.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our