Image Credit: Shutterstock
The White House has accused Chinese developer Moonshot AI of using outputs from Anthropic’s Claude Fable model to help develop Kimi K3, an allegation the administration has not publicly substantiated.
Michael Kratsios, director of the White House Office of Science and Technology Policy, said on July 22 that the administration had information linking the new model to large-scale covert distillation. He did not release technical evidence supporting the allegation.
Kratsios Alleges Moonshot Covertly Accessed U.S. AI Models
Kratsios said Moonshot built an internal platform that could switch between multiple access methods while extracting capabilities from U.S. models.
He distinguished routine distillation, which developers use to make smaller or more efficient systems, from covert activity intended to copy proprietary technology. The administration has not announced charges, sanctions, or export-control penalties against Moonshot.
Treasury Secretary Scott Bessent has said Chinese AI companies could face sanctions if improper distillation or intellectual property theft is confirmed. Moonshot had not issued a public response to the Kimi K3 accusation when it was reported.
Anthropic Linked Moonshot to 3.4M Claude Exchanges
Anthropic separately accused Moonshot in February of conducting more than 3.4 million exchanges with Claude through hundreds of fraudulent accounts. The company said the traffic targeted agentic reasoning, coding, data analysis, computer use and vision capabilities.
Anthropic attributed the campaign to Moonshot through request metadata that it said matched public profiles of senior employees. It also alleged that a later phase attempted to reconstruct Claude’s reasoning traces.
Anthropic Findings Do Not Prove Kimi K3 Used Claude Data
Anthropic’s February findings describe activity the company associated with Kimi models generally. They do not publicly show which data, if any, entered Kimi K3’s training process.
That distinction leaves the White House allegation separate from the earlier Anthropic disclosure until technical evidence or enforcement filings are released. The key unresolved issue is whether officials can link Claude-derived outputs directly to Kimi K3 development.
Kimi K3 Launches With 2.8 Trillion Parameters
Moonshot released Kimi K3 on July 16 as a 2.8-trillion-parameter, natively multimodal model with a one-million-token context window. The company markets it for coding, knowledge work, and long-horizon reasoning.
The model’s scale has intensified scrutiny, but it does not prove the White House allegation. Kimi K3 has performed strongly in independent coding evaluations and has been presented as a lower-cost competitor to leading systems from Anthropic and OpenAI.
White House Has Not Announced Sanctions or Charges
Anthropic’s public policy chief supported Kratsios’ statement and said illicit distillation created intellectual property and national security risks. The release renewed debate over Chinese open-weight models and U.S. controls on advanced AI technology.
The next confirmed step depends on whether the administration publishes evidence or takes formal action against Moonshot. Until then, the accusation remains a high-stakes claim without public technical proof.
