OpenAI, Anthropic accuse Chinese rivals of AI data theft

25 FEBRUARY 2026

AI firm Anthropic said it uncovered campaigns by three China AI firms to illicitly extract capabilities from Claude chatbot, in what it described as industrial-scale intellectual property theft.
Anthropic said DeepSeek, Moonshot AI and MiniMax used “distillation” to rapidly boost the performance of a less capable one.
Rival OpenAI, creator of ChatGPT, made similar accusations to U.S. lawmakers earlier.

Anthropic, the company behind the Claude chatbot, claimed that three Chinese AI firms tried to illegally extract capabilities from its system.
It described the activity as “industrial-scale intellectual property theft.”
The goal, according to Anthropic, was to copy or replicate Claude’s advanced capabilities.
OpenAI (creator of ChatGPT) reportedly made similar accusations.
It told U.S. lawmakers that some foreign AI firms were using ChatGPT outputs to train their own competing models
Three Chinese AI companies, DeepSeek, Moonshot AI and MiniMax allegedly used a technique called distillation.

“Distillation” in AI

Model distillation is when a smaller or less capable AI model is trained using the outputs of a larger, more powerful model.
The smaller model “learns” by observing responses from the bigger model.
This can significantly improve the smaller model’s performance.

Anthropic itself accused

Earlier, Anthropic itself was accused by authors and content creators of using copyrighted books and online material without permission to train its AI model, Claude.
Several writers alleged that Anthropic used pirated or copyrighted books for AI training.
A group of authors filed lawsuits claiming their works were used without consent.
The issue is part of a broader legal debate over whether AI companies can train models on publicly available or copyrighted data without explicit permission.