A model that was celebrated as a breakthrough in efficient AI training just five days after launch is now at the center of a U.S.-China trade and technology dispute. Michael Kratsios, Directo
A model that was celebrated as a breakthrough in efficient AI training just five days after launch is now at the center of a U.S.-China trade and technology dispute.
Michael Kratsios, Director of the White House Office of Science and Technology Policy, publicly accused Chinese AI company Moonshot AI of accessing banned Nvidia chips to train Kimi K3, its 2.8-trillion-parameter open-weight model.
According to the accusation, Moonshot didn’t import restricted Nvidia GB300 chips into China directly. Instead, the company reportedly accessed GB300-equipped servers through infrastructure located in Thailand — training on the hardware remotely without ever bringing the physical chips across a restricted border, according to TFTC’s reporting.
Kratsios also alleged Moonshot ran an industrial-scale covert distillation operation against Anthropic’s Fable model, using a system designed to rotate access methods and avoid detection.
Nvidia shares fell sharply on the news as investors weighed the implications for export-control enforcement broadly.
Kimi K3 was released as an open-weight model on July 16 — meaning its trained parameters are already downloaded, mirrored, and running on servers around the world. Even if U.S. regulators pursue sanctions against Moonshot directly, they cannot recall model weights that are already public. That distinction matters enormously for how effective any resulting policy response can actually be.
This is the second time in 18 months that a low-cost Chinese model has rattled U.S. AI markets — DeepSeek’s R1 did the same in January 2025, wiping roughly $600 billion off Nvidia’s market cap in a single day. What’s different this time is the explicit allegation of sanctions evasion rather than just efficient training methods, which raises the stakes for export-control policy and for any company relying on the current enforcement regime to protect its competitive position.
The episode is also arriving days after Nvidia CEO Jensen Huang publicly argued against restricting open-weight models — creating a visible tension between Nvidia’s commercial interest in maximizing chip demand and Washington’s national-security concerns.
Analysts are divided on how much of Kimi K3’s leap in capability is genuinely attributable to distillation versus fundamental efficiency research. Moonshot has previously said publicly that export controls forced it to focus on training efficiency rather than simply scaling up compute — a narrative that predates the current allegations and complicates any simple story about the model’s origins.
Some industry voices have also pointed out the political dimension: U.S. companies openly debating whether to adopt Chinese open-weight models is itself unusual, and reflects how competitive Kimi K3’s pricing and benchmark performance have become.
The post Kimi K3 Fallout: White House Accuses Moonshot AI of Evading Nvidia Chip Export Controls appeared first on Times Tabloid.
