BotBeat
...
← Back

> ▌

AnthropicAnthropic
POLICY & REGULATIONAnthropic2026-07-24

White House Accuses China's Moonshot AI of Stealing Anthropic's Fable Model via Distillation

Key Takeaways

  • ▸White House officials accuse Moonshot AI of using distillation to steal Anthropic's Fable model and create the Kimi K3 model released July 16, 2026
  • ▸The alleged theft involved accessing restricted NVIDIA GB300 GPUs running in Thailand and developing an internal platform to evade detection
  • ▸US Treasury Secretary warns of potential sanctions and Entity List designations if China continues large-scale IP theft through distillation
Source:
Hacker Newshttps://www.theregister.com/ai-and-ml/2026/07/23/senior-white-house-official-claims-chinas-k3-model-stolen-from-anthropic/5276804↗

Summary

Michael Kratsios, Trump's Assistant for Science and Technology, publicly accused China's Moonshot AI of distilling Anthropic's Fable model to create its Kimi K3 model, which was released on July 16, 2026. Kratsios alleged that Moonshot AI built a "sophisticated internal platform" to conduct large-scale distillation attacks against U.S. models and utilized NVIDIA GB300 accelerators running in Thailand—hardware the U.S. restricts from export to China. The K3 model's impressive 2.8-trillion-parameter performance sparked significant investor concern about China's AI capabilities, contributing to declines in U.S. AI stock valuations.

US Treasury Secretary Scott Bessent reinforced the accusations, warning that while the U.S. "supports open-source AI and the innovation it unlocks," covert industrial-scale distillation constitutes IP theft and will trigger "sanctions and Entity List designations." Anthropic had previously accused Moonshot AI of similar distillation practices in February 2026, alongside allegations against other Chinese competitors DeepSeek and MiniMax. The U.S. has extensively sanctioned China's tech sector, though these measures have not prevented Chinese firms from developing sophisticated workarounds, including renting GPU farms outside China to access restricted hardware.

  • Anthropic had previously accused Moonshot AI and other Chinese firms (DeepSeek, MiniMax) of similar distillation attacks in February 2026
Large Language Models (LLMs)Generative AIGovernment & DefenseRegulation & Policy

More from Anthropic

AnthropicAnthropic
PRODUCT LAUNCH

Anthropic Releases Claude Opus 5: Mid-Tier Model Balances Performance and Affordability

2026-07-25
AnthropicAnthropic
RESEARCH

Dragos: Real-World Cyberattack Used Claude and GPT to Breach Water Utility OT Systems

2026-07-25
AnthropicAnthropic
INDUSTRY REPORT

Silicon Valley Splits Over Chinese AI: Safety vs. Access Debate Intensifies

2026-07-25

Comments

Suggested

CloudflareCloudflare
UPDATE

Cloudflare Expands AI Bot Controls With Nuanced Classification System

2026-07-25
AnthropicAnthropic
PRODUCT LAUNCH

Anthropic Releases Claude Opus 5: Mid-Tier Model Balances Performance and Affordability

2026-07-25
OpenAIOpenAI
POLICY & REGULATION

OpenAI's AI Models Break Free: First Real Loss-of-Control Incident Exposes Regulatory Gaps

2026-07-25
← Back to news
© 2026 BotBeat
AboutPrivacy PolicyTerms of ServiceContact Us