Anthropic, a U.S.-based AI company, revealed that several Chinese AI labs, including Alibaba, Moonshot, and DeepSeek, engaged in unauthorized activity to train their models by using outputs from Anthropic’s Claude AI. This practice, known as “illicit distillation,” involves extracting responses from a more advanced AI system and utilizing them to improve other AI models without consent. Anthropic said this occurred on a large scale, with millions of exchanges secretly harvested from Claude without proper authorization.

The report detailed Alibaba’s involvement as the most extensive, with over 151 million exchanges recorded between May and July 2026. Alibaba allegedly used these exchanges to train its Qwen AI models and for other research purposes such as reinforcement learning and model architecture development. Moonshot, a Beijing-based company behind the Kimi models, reportedly routed nearly 300,000 customer requests intended for their own AI to Claude, misleading users into thinking the replies came from Kimi. Moonshot saved and repurposed many exchanges to train its models, affecting over 23 million exchanges from May to July.

DeepSeek also participated in similar unauthorized practices, with more than 12 million distillation events detected in a two-week period during July 2026. These activities raised concerns not only about intellectual property and competitive fairness but also about potential breaches of user privacy and legal terms, as some exchanged data contained sensitive information from individuals, businesses, and government-affiliated entities. Anthropic emphasized that such practices likely violated privacy laws and the companies’ own service agreements.

Anthropic’s threat intelligence report covers a broader timeframe from December 2025 to August 2026 and highlights misuse of Claude AI across various domains including cybersecurity, disinformation, surveillance, and fraud. While the report named multiple Chinese AI companies, Alibaba, Moonshot, DeepSeek, Xiaomi, and Anthropic did not provide immediate comments to CNBC. The allegations underscore growing tensions in the AI industry about ethical use of proprietary AI technologies and global competition in advanced artificial intelligence innovation.