The White House has accused Chinese AI startup Moonshot AI of secretly using technology from Anthropic to develop its Kimi K3 model, which launched on July 16, 2026. The allegations were made public by Michael Kratsios, director of the White House Office of Science and Technology Policy, in a post on X on Wednesday.

"We have information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model," Kratsios wrote. "To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of…"

— Director Michael Kratsios (@mkratsios47) July 22, 2026

Kratsios alleged that Moonshot AI built a covert internal platform designed to carry out large-scale distillation of US AI models. The platform reportedly enabled the company to switch between access methods to avoid detection. He also claimed that Moonshot AI expanded its computing capacity by acquiring servers in Thailand to train its models.

Distillation involves training one AI model on the outputs of another, allowing a developer to replicate some of a frontier model's capabilities without access to its underlying weights or training data. The practice has become a growing concern for US AI labs that invest billions in developing frontier models, as it can allow competitors to shortcut that investment.

Kimi K3 has been widely praised as one of China's most capable AI models since its release. The accusations underscore Washington's growing concern that Chinese firms are leveraging American AI models to accelerate their own development, adding another front to the technology competition between the two countries that has already seen US restrictions on advanced chip exports to China.

Researchers Question the Timeline

Some AI researchers have pushed back against the White House's claims, arguing that the timeline between Anthropic's model release and Kimi K3's launch is too compressed for distillation to explain the model's performance.

Anthropic's Fable 5 model was re-released on July 1 after export restrictions were lifted. Kimi K3 launched just 15 days later, on July 16.

Elie Bakouch, a researcher at AI startup Prime Intellect, said that timeframe makes distillation unlikely. "I don't think claiming that K3's performance comes from Fable distillation makes sense technically," he said.

Dean Ball, OpenAI's head of strategic futures, also weighed in, stating that he did not believe Kimi K3's performance could be explained by distillation alone.

The White House has not released public evidence to substantiate its claims. Moonshot AI had not responded publicly at the time of writing.

Sanctions and Export Restrictions on the Table

US Treasury Secretary Scott Bessent signaled that the administration views large-scale distillation as a serious threat to American intellectual property.

"Open source is not open season on American IP," Bessent said.

He warned that Chinese firms conducting covert distillation at scale could face sanctions and placement on the Entity List, a trade restriction mechanism that limits access to US technology. Companies added to the Entity List face stringent licensing requirements for any US-origin technology, with most license applications subject to a presumption of denial.

Kratsios drew a distinction between legitimate distillation — a well-established technique for creating smaller, more efficient models — and what he characterized as the theft of US intellectual property.

Anthropic confirmed earlier in July that the US Department of Commerce had lifted export restrictions on its most powerful models, including Fable 5 and Mythos 5. The company said it reopened international access the following day.

Both Moonshot AI and Anthropic have been eyeing initial public offerings, adding a financial dimension to the dispute. The US government's response to the alleged activity remains ongoing.