Anthropic said on September 10 that it detected and disrupted what it described as unauthorized efforts by China-based AI labs, including Alibaba, Moonshot AI, and DeepSeek, to train models using Claude outputs.

Anthropic also alleged that some companies routed user prompts to Claude without notifying users.

Anthropic said the Alibaba-linked operation was the largest distillation campaign it had measured, with more than 151 million exchanges with Claude from May to July. It attributed more than 23 million exchanges over the same period to Moonshot AI, a Beijing-based developer of the Kimi family of AI models.

Anthropic said it saw more than 12 million distillation attacks attributable to DeepSeek over 14 days in July 2026. Some customer requests sent to Claude contained sensitive information.

The allegations add to broader industry concern about the use of large language model outputs to distill rival models.

🔗 Source: CNBC

Recent Anthropic developments