Kimi K3 Scandal: White House Accuses Moonshot of Stealing American AI Technologies
The technological confrontation between the US and China in the field of artificial intelligence has reached a new boiling point. White House AI advisor Michael Kratsios has leveled direct accusations against the Chinese startup Moonshot AI. According to him, the company secretly copied Anthropic's Fable 5 model and used it to create its own flagship development — Kimi K3.
This involves the so-called "distillation" — a technique where a new model is trained on the responses of a more powerful AI. In an open form, this is a legal and common practice. However, as Kratsios emphasizes, covert distillation on an industrial scale aimed at stealing American technology is no longer innovation, but theft.
Large-Scale Attack on Claude
According to an internal report from Anthropic, the company's specialists recorded over 3.4 million suspicious requests to the Claude model linked to Moonshot. The traffic came from hundreds of fake accounts, and the account data matched the public profiles of senior employees at the Chinese startup. The target of the attack was Claude's key capabilities: logical reasoning, code writing, and image processing.
Kratsios also stated that Moonshot acquired NVIDIA GB300 servers — top-tier AI chips whose export to China is strictly prohibited by US regulators. The equipment was allegedly deployed in Thailand. It is worth recalling that in June, Super Micro's office in Taiwan was already raided amid suspicions of chip smuggling to China.
Kimi K3: Record Launch and Security Risks
Moonshot unveiled the Kimi K3 model on July 16. With 2.8 trillion parameters, it is the largest open-source model ever released. Demand was so high that the company suspended new subscriptions within 48 hours.
However, the publication date is critically important. Anthropic warns that copied models lose their built-in safety restrictions. Once released into the public domain, such versions can spread uncontrollably, posing a direct threat.
This is not the first such incident. In January 2025, White House AI chief David Sacks accused DeepSeek of copying OpenAI's models, although Beijing dismissed these conclusions as unsubstantiated. For now, Kratsios has not announced specific sanctions, and Moonshot is refraining from comments.
My analysis: The Moonshot situation is not just a corporate conflict, but a symptom of a systemic problem. The US maintains a huge advantage in AI investment (spending nearly 23 times higher than China's), but Chinese developers are rapidly closing the gap through aggressive technology borrowing. The next move is up to Washington, and it could be harsh.