Home / Technology / Alibaba Accused of Massive AI Model Theft
Alibaba Accused of Massive AI Model Theft
25 Jun
Summary
- Anthropic alleges Alibaba illicitly extracted Claude AI model capabilities.
- A large-scale distillation attack generated over 28.8 million exchanges.
- The campaign aimed to boost China's AI development with stolen data.
Anthropic, a prominent U.S. artificial intelligence company, has formally accused Chinese e-commerce and technology conglomerate Alibaba of illicitly extracting capabilities from its Claude AI model. This accusation stems from what Anthropic describes as the most significant distillation attack ever perpetrated against the company.
Distillation attacks involve training a less advanced model on the outputs of a more capable one to steal its proprietary abilities. Anthropic reported that this specific campaign operated between April 22 and June 5, 2026. It utilized almost 25,000 fraudulent accounts to generate over 28.8 million interactions with Claude.
The AI firm asserted in a letter that such operations serve as a method for competitors to obtain advanced American AI technology, thereby accelerating China's progress towards matching its most sophisticated models. The letter indicated that operators affiliated with Alibaba and its AI division, Alibaba Qwen, were behind this alleged campaign.
This communication, dated June 10, 2026, was directed to U.S. Senators Tim Scott and Elizabeth Warren, key members of the Senate Banking Committee, in anticipation of a scheduled hearing focused on artificial intelligence.
