White House technology official Michael Kratsios accused China’s Moonshot AI of using Anthropic’s Fable model to help develop Kimi K3, according to Reuters and CyberScoop. Anthropic has separately alleged large-scale extraction attempts involving Moonshot-linked activity, while Moonshot’s own Kimi K3 post describes...
White House technology official Michael Kratsios accused Beijing-based Moonshot AI of using Anthropic’s Fable model to help develop its Kimi K3 system, according to Reuters via Investing.com and CyberScoop.
Reuters reported that Kratsios said Moonshot “distilled” Anthropic’s Fable model for its latest AI system, Kimi K3. CyberScoop also reported the accusation, noting that Kratsios alleged Moonshot used Anthropic’s model to develop K3 but did not provide details about the U.S. government’s evidence.
Distillation is a common machine-learning technique in which one model is trained to imitate outputs or behavior from another model. In legitimate settings, it can be used to make smaller or more efficient models. The allegation at issue is not ordinary research reuse, but whether a company covertly extracted proprietary model behavior at scale.
Reuters said Moonshot and the Chinese Embassy in Washington did not immediately comment on the allegation. CyberScoop similarly framed the accusation as a White House claim rather than an independently established fact.
Anthropic has previously described what it calls “distillation attacks” against its Claude models. In a February company post titled “Detecting and preventing distillation attacks,” Anthropic said it had identified attempts to use fraudulent accounts to extract model capabilities.
Anthropic’s post directly named Moonshot among organizations it said were involved in suspicious activity. The company alleged that activity linked to Moonshot included more than 3.4 million exchanges with Claude. Anthropic presented the activity as an example of large-scale extraction designed to reproduce another model’s capabilities, rather than normal use of a commercial AI service.
The company said such behavior can harm model developers by allowing competitors to copy capabilities without incurring comparable research and training costs. Anthropic also said it had developed detection and prevention methods, including monitoring for abnormal usage patterns.
Moonshot’s own Kimi K3 announcement describes the model as a large “open frontier intelligence” system. The company says Kimi K3 has 2.8 trillion parameters, native vision capabilities, and a 1 million-token context window. Moonshot also says it plans to release full model weights by July 27, 2026.
The Kimi post focuses on the model’s technical specifications and release plans. It does not address the White House allegation reported by Reuters and CyberScoop, nor Anthropic’s earlier claims about distillation attacks.
The dispute highlights a growing policy and business conflict around model access. Frontier AI companies often expose their models through paid APIs, which makes them useful to customers and developers but can also create opportunities for competitors or automated systems to generate large volumes of training data from model outputs.
For AI labs, the line between acceptable learning from public research and improper extraction from proprietary systems is becoming increasingly important. Research papers, open-source code, public datasets, and model outputs all play different roles in AI development, but companies are now pressing for clearer boundaries around high-volume imitation of commercial models.
The White House accusation also lands amid broader U.S.-China competition over advanced AI and compute access. Reuters reported Kratsios’s comments in the context of concerns that Chinese AI firms may be using U.S.-developed models or infrastructure to accelerate their own systems. CyberScoop noted that the official did not disclose the underlying evidence, leaving key questions unanswered publicly.
For now, the public record supports a narrower conclusion: a senior White House technology official has accused Moonshot of distilling Anthropic’s Fable model for Kimi K3; Anthropic has separately alleged large-scale extraction activity involving Moonshot; and Moonshot’s public Kimi K3 materials describe the model’s capabilities without responding to those claims.
White House technology official Michael Kratsios accused Beijing based Moonshot AI of using Anthropic’s Fable model to help develop its Kimi K3 system, according to Reuters via Investing.com and CyberScoop.
What the White House alleged Reuters reported that Kratsios said Moonshot “distilled” Anthropic’s Fable model for its latest AI system, Kimi K3.
CyberScoop also reported the accusation, noting that Kratsios alleged Moonshot used Anthropic’s model to develop K3 but did not provide details about the U.S.
Continue reading