Background
<cite index="2-2,2-3">On July 22, Michael Kratsios, the White House's director of science and technology policy, accused Beijing-based Moonshot AI of conducting large-scale, covert distillation of Anthropic's Fable model to build its Kimi K3 system — the first time a senior U.S. official has publicly named a specific Chinese laboratory for copying a specific American model.</cite>
The Accusation
Model distillation is a machine-learning technique in which a smaller "student" model is trained on the outputs of a larger "teacher" model to replicate its capabilities at reduced computational cost. <cite index="5-7,5-8">While this process can infringe on intellectual property rights, it is also widely used as a legitimate optimization method.</cite>
<cite index="3-3">Kratsios alleged that Moonshot developed a sophisticated internal platform to conduct large-scale distillation against U.S. models, using methods designed to evade detection.</cite> <cite index="4-4">He also said the company acquired Nvidia GB300 servers — either directly or through Thailand — to train its artificial intelligence (AI) models.</cite> <cite index="5-4">The GB300 servers are part of Nvidia's Blackwell generation, which are banned from being sold to Chinese companies.</cite>
<cite index="3-4,3-5">Kratsios stated that "legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem," but added that "large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable."</cite>
Treasury Warning
<cite index="3-6,3-7">U.S. Treasury Secretary Scott Bessent warned that large-scale distillation attacks could result in sanctions and other restrictions, stating that "when PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table."</cite> <cite index="2-13,2-14">As of July 23, no formal sanctions or Entity List designations had been announced, though the simultaneous public statements from the White House science director and the Treasury Secretary signaled the administration was building a case for action.</cite>
Kimi K3 and Prior Allegations
<cite index="4-9">Moonshot AI describes Kimi K3 as the first open 2.8-trillion-parameter model, and promotes its lower token costs while still delivering near-frontier performance.</cite> <cite index="2-8,2-9">Anthropic's Fable model became publicly available on July 1, 2026, and Moonshot released Kimi K3 as an open-weight model the following week.</cite> <cite index="2-12">The full Kimi K3 weights are scheduled to go public on July 27, a release that would make the model's architecture and training methods available for independent scrutiny.</cite>
The White House accusation builds on a prior disclosure by Anthropic. <cite index="19-4,19-5">In February 2026, Anthropic accused three prominent Chinese AI labs — DeepSeek, MiniMax, and Moonshot AI — of illegally extracting capabilities from its Claude model, alleging they created over 24,000 fraudulent accounts and trained their models using over 16 million exchanges with Claude.</cite> <cite index="22-9">Moonshot's share of that activity amounted to approximately 3.4 million exchanges, focused on agentic reasoning and computer-use agent development.</cite> <cite index="18-8">Despite Anthropic's service restrictions preventing commercial access to Claude in China, the firms allegedly engaged commercial proxy services to sidestep those restrictions.</cite>
Disputed Claims and Open Questions
<cite index="2-10">The compressed timeline between Fable's public launch and K3's release has fuelled suspicion that K3's capabilities were bootstrapped from American technology, though several independent AI researchers have questioned whether a model released just days after Fable's public launch could have been primarily built through distillation.</cite> <cite index="4-7">Kratsios did not provide details on how the U.S. government learned that K3 had been distilled from Anthropic's model.</cite> <cite index="2-11">As of July 23, Moonshot AI had not publicly responded to the accusations.</cite>
<cite index="12-9,12-10,12-11">Founded in 2023 by Yang Zhilin, a Tsinghua University graduate who previously conducted research at Google and Meta, Moonshot AI quickly became one of China's most prominent AI startups, gaining early traction in 2024, and had raised roughly $1.5 billion across multiple rounds by early 2026 with its valuation reaching $4.3 billion.</cite>