US artificial intelligence company Anthropic accused Chinese technology giant Alibaba of illicitly accessing its Claude AI model through a large-scale campaign aimed at copying its advanced capabilities, according to the Wall Street Journal on Thursday.
In a June 10 letter sent to US senators Tim Scott and Elizabeth Warren, Anthropic said operators linked to Alibaba and its Qwen AI lab carried out what it described as the "largest known distillation attack" against the company to date, according to the report.
The campaign involved nearly 25,000 fraudulent accounts and 28.8 million exchanges with Claude between April 22 and June 5.
Anthropic said the activity targeted some of Claude's most advanced capabilities, including software engineering, agentic reasoning and long-horizon task planning.
The company argued that such campaigns allow rival AI developers to copy the capabilities of leading US models without bearing the high research, development and training costs required to build frontier systems.
Anthropic also urged US lawmakers and officials to take stronger action against unauthorized model-copying efforts, including restrictions on access to advanced US computing infrastructure and penalties against entities involved in such campaigns.
In artificial intelligence, distillation generally refers to training a smaller or less capable model using the outputs of a more advanced system. While the technique can be used for legitimate model development, major AI labs have warned that unauthorized use of frontier models for this purpose may violate terms of service and raise intellectual property and national security concerns.
The allegations come as competition between the US and China in artificial intelligence intensifies, with Washington tightening export controls on advanced chips and AI-related technologies while Chinese companies continue to expand their own large language model offerings.
