The US government has accused Chinese AI company Moonshot AI of using Anthropic’s Fable model to help develop Kimi K3, its latest large language model.
Michael Kratsios, a senior technology official in the Trump administration, said US authorities had information showing that Moonshot AI distilled Fable during K3’s development. He described the alleged activity as a large-scale attempt to take proprietary American technology.
The claims have raised fresh concerns about intellectual property theft, AI competition and national security. Meanwhile, US officials are considering whether to add Moonshot AI to a trade blacklist and impose sanctions.
US Claims Moonshot AI Used Anthropic’s Fable
Kratsios said Moonshot AI allegedly used outputs from Anthropic’s Fable model to improve Kimi K3.
AI distillation involves training one model on the responses produced by another. Developers often use the method to build cheaper and smaller systems. However, the practice can become controversial when companies access a competitor’s model without permission.
According to Kratsios, Moonshot AI used distillation to obtain capabilities developed by Anthropic. He said US officials view such activity as an attempt to undermine American AI research.
Moonshot AI did not immediately comment on the accusations.
However, Liu Chang, a spokesperson for the Chinese embassy in Washington, rejected the claims. He said the allegations lacked evidence and added that China respects intellectual property rights.
Treasury Considers Sanctions and Trade Blacklist
US Treasury Secretary Scott Bessent said the government is weighing possible action against Moonshot AI.
Potential measures include adding the company to a trade blacklist and imposing sanctions. Such restrictions could make it harder for Moonshot AI to access US technology, software and business partners.
The move would add to growing technology tensions between the United States and China. Both countries are investing heavily in advanced AI systems, which governments increasingly view as important for economic growth, cybersecurity and military development.
Anthropic public policy executive Sarah Heck also warned that alleged theft of US AI models could create national security risks for the country.
Moonshot AI Also Faces Nvidia Chip Questions
Kratsios further claimed that Moonshot AI obtained servers containing Nvidia GB300 AI chips and used them in Thailand.
He said the company likely used the equipment to train AI models. The allegations could draw further scrutiny because US export controls limit access to advanced AI chips in China.
The US government has previously raised similar concerns about other Chinese AI companies. In February, a senior Trump administration official claimed that DeepSeek trained a recent model using Nvidia’s most advanced chip technology.
US authorities have not provided public evidence to support the latest claims involving Moonshot AI.
Anthropic Previously Reported Suspected Abuse
Anthropic said in February that three Chinese AI companies had improperly used Claude to improve their own models.
The company named DeepSeek, Moonshot AI and MiniMax. According to Anthropic, the firms created more than 16 million interactions with Claude through around 24,000 fake accounts.
Anthropic said the activity breached its terms of service and bypassed regional access restrictions.
OpenAI has also warned US lawmakers that Chinese competitors may try to replicate leading American models. The company reportedly raised concerns that DeepSeek targeted US AI firms to collect outputs for model training.
State Department Warned of AI Distillation Campaigns
The US State Department reportedly told diplomats in April to highlight alleged intellectual property theft by Chinese AI companies.
A diplomatic cable warned that unauthorised AI distillation can help foreign companies create products that appear competitive on selected benchmarks at a much lower cost.
However, the cable also noted that those systems may not match the original model’s full capabilities. A model can imitate answers in specific tests without reproducing the broader reliability, safety work and underlying performance of the original system.
Kimi K3 Launches Amid Growing AI Rivalry
Moonshot AI launched Kimi K3 shortly before the allegations emerged. The company described it as a 2.8 trillion-parameter open-weight model and said it performs close to Anthropic’s Fable system.
The launch shows how quickly Chinese AI companies are advancing, particularly in open-weight models. Yet the Moonshot AI distillation allegations may now place Kimi K3 under greater pressure from US officials.
If Washington moves forward with sanctions or blacklist measures, Moonshot AI could face more difficulty accessing critical AI hardware and international technology partners.
0 responses to “US Accuses Moonshot AI of Distilling Anthropic’s Fable Model”