USA accuses China’s Moonshot AI of stealing Anthropic’s Fable to build Kimi 3 model

The United States has accused Chinese AI startup Moonshot AI of illegally using Anthropic’s proprietary Fable model to develop its latest open-weight model, Kimi K3, escalating tensions over artificial intelligence, intellectual property, and U.S.-China technology competition. The allegations were made by Michael Kratsios, Director of the White House Office of Science and Technology Policy, who claimed Moonshot conducted large-scale “covert distillation” of Anthropic’s most advanced model to build K3 while attempting to evade detection.

The accusations represent one of the strongest public statements yet by the Trump administration against a Chinese AI company. While model distillation—the process of using one AI model to help train another—is a common AI development technique, U.S. officials argue that using proprietary models without authorization amounts to theft of intellectual property. Moonshot AI has not publicly responded to the latest allegations.

White House Alleges Covert Use of Anthropic’s Fable

According to Michael Kratsios, Moonshot AI allegedly:

• Used Anthropic’s Fable model without authorization.
• Conducted large-scale AI model distillation.
• Built an internal platform designed to avoid detection while querying U.S. AI systems.
• Leveraged the extracted knowledge to develop its Kimi K3 model.

Kratsios emphasized that while distillation itself is a legitimate machine learning technique, secretly extracting capabilities from proprietary frontier models is “unacceptable” when it involves unauthorized access to American technology.

Allegations at a Glance

Aspect	Details
Accuser	White House Office of Science and Technology Policy
Company Accused	Moonshot AI (China)
Alleged Source Model	Anthropic Fable
Target Model	Kimi K3
Main Allegation	Unauthorized large-scale model distillation

What Is Model Distillation?

Model distillation is a widely used AI training technique in which a smaller or newer model learns from the outputs of a more capable model.

Legitimate uses typically involve:

• Training with the model owner’s permission.
• Internal optimization.
• Knowledge transfer within an organization.

The controversy arises when companies allegedly use proprietary commercial models without authorization to improve competing systems. U.S. officials argue that this crosses the line from engineering practice into intellectual property infringement.

Why Kimi K3 Has Drawn Attention

Moonshot AI’s Kimi K3 has attracted significant industry attention after delivering performance comparable to leading frontier AI models while being released as an open-weight model.

The model is known for:

• Strong reasoning capabilities.
• Competitive coding performance.
• Open-weight availability.
• Lower deployment costs compared with many proprietary alternatives.

Its rapid rise has intensified scrutiny from U.S. policymakers concerned about the pace of Chinese AI development.

Key Players

Organization	Role
Moonshot AI	Developer of Kimi K3
Anthropic	Developer of the proprietary Fable model
White House OSTP	Publicly announced the allegations
U.S. Treasury	Indicated sanctions remain under consideration if violations are confirmed

Anthropic Supports Stronger Protection

Anthropic has backed the administration’s tougher stance on unauthorized model distillation.

The company argues that:

• Large-scale extraction of proprietary model outputs threatens AI innovation.
• Frontier AI models require stronger intellectual property protections.
• Unauthorized distillation could undermine incentives for developing advanced AI systems.

Potential Geopolitical Impact

The allegations come amid growing competition between the United States and China over artificial intelligence leadership.

Potential implications include:

• Increased scrutiny of Chinese AI companies.
• Possible sanctions if evidence supports the allegations.
• Greater emphasis on AI intellectual property protection.
• Tighter controls over access to frontier AI models.
• Further escalation in U.S.-China technology tensions.

No Public Evidence Released Yet

While U.S. officials have publicly accused Moonshot AI of unauthorized distillation, they have not released detailed technical evidence supporting the allegations.

At the time of writing:

• No formal legal ruling has been issued.
• No public forensic report has been released.
• Moonshot AI has not publicly responded to the latest accusations.
• The claims remain allegations by U.S. officials rather than established legal findings.

Looking Ahead

The accusations against Moonshot AI mark a significant escalation in the debate over AI intellectual property and model distillation. As frontier AI models become increasingly valuable strategic assets, governments and developers are placing greater emphasis on protecting proprietary technologies from unauthorized use. While distillation remains a legitimate technique in many contexts, the White House argues that covert use of commercial models without permission constitutes theft rather than innovation.

Looking ahead, the dispute could influence future AI regulation, export controls, and enforcement of model usage policies. Whether additional evidence is released—or whether regulatory or legal action follows—will likely shape how AI companies worldwide approach model training, licensing, and intellectual property protection in an increasingly competitive global AI landscape.