The White House is weighing how to respond to increasingly capable Chinese AI models, with officials divided over whether new restrictions would be effective or workable. WIRED reports that the debate has intensified after Moonshot AI released Kimi K3, a model described as competitive with leading US systems.
According to the report, parts of the White House favor stricter controls, while the Commerce Department has been more skeptical because it oversees export controls and would have to turn policy into enforceable rules. One focus is distillation, the practice of training one model on another model’s outputs.
The issue became more urgent after Anthropic alleged that Alibaba carried out a large distillation attack against Claude. Treasury Secretary Scott Bessent has described covert distillation as intellectual-property theft that could lead to sanctions.
No final policy has been announced. The practical difficulty is that AI model weights, APIs, open releases, and research exchanges do not fit neatly into traditional export-control categories, even as Washington grows more concerned about the strategic value of frontier models.