Moonshot AI Faces US Accusations of Distilling Anthropic Technology

Moonshot AI’s Kimi K3 and the Technical Controversy

The model, which features a massive parameter count and a mixture-of-experts architecture, has drawn significant attention for its performance in complex reasoning and front-end coding. Demand was so immediate that the company was forced to temporarily halt new subscriptions within 48 hours of the launch.

However, the technical achievement is overshadowed by serious allegations from the U.S. government. Michael Kratsios, Director of the Office of Science and Technology Policy at the White House, publicly accused Moonshot AI of using industrial, hidden and large-scale distillation to copy technology from Anthropic’s Claude models. This technique involves training a new AI using the outputs of an existing, more advanced model.

Michael Kratsios, Director of the Office of Science and Technology Policy at the White House, stated that they have information indicating Moonshot AI distilled Anthropic’s Fable model for the development of its K3 model, adding that industrial, hidden, and large-scale distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable.

Anthropic claims to have tracked over 3.4 million interactions from accounts linked to Moonshot employees directed at Claude. Moonshot AI has denied these accusations. The tension is further complicated by allegations that the company utilized restricted NVIDIA GB300 chips in Thailand, circumventing U.S. export controls.

U.S. Government Pressure on Taiwan Engagement

As the technological rivalry intensifies, the U.S. government is simultaneously pushing back against Chinese diplomatic interference. On June 25, the U.S. Departments of State, Agriculture, and Commerce released joint letters addressed to U.S. governors and business leaders. These documents reveal that Chinese embassies and consulates have been actively contacting American officials and firms to discourage engagement with Taiwan.

The letters assert that Beijing has been mischaracterizing U.S. policy by, for example, falsely claiming that Washington has previously accepted Beijing’s specific position on Taiwan. The U.S. government advised that any entity receiving such pressure should contact the State Department directly.

In Beijing, Chinese foreign ministry spokesperson Guo Jiakun responded to the move, stating that China consistently and firmly opposes any form of official interaction between the U.S. and Taiwan. Taiwan’s government, however, welcomed the U.S. show of support, noting that the letters reaffirm Taiwan’s role as a vital partner in the global trade system.

Market Impact and Future Uncertainties

The release of Kimi K3 has reignited debates regarding the effectiveness of U.S. export restrictions. While Washington has sought to limit China’s access to advanced chips and AI research, the rapid development of open-source models like K3 suggests that Chinese firms are finding ways to remain competitive.

The industry is now watching for the next phase of this competition. The full weights of the Kimi K3 model are scheduled to be made public on July 27. Security experts, including those at Anthropic, have warned that the transition to an open-source model could lead to the loss of safety guardrails, allowing the technology to proliferate without the constraints typically applied to proprietary, closed-source models.

Whether these models will continue to close the gap with U.S. leaders like OpenAI and Anthropic remains a point of contention. While Kimi K3 has performed strongly in specialized benchmarks—even surpassing some versions of Claude and GPT in programming tasks—it remains to be seen if it can maintain this trajectory under the weight of international scrutiny and ongoing trade restrictions.