Artificial intelligence company Anthropic has accused Chinese technology giant Alibaba of carrying out what it describes as the largest known attempt to extract capabilities from its Claude AI models. According to a letter sent to U.S. lawmakers, Anthropic claims that operators affiliated with Alibaba and its AI division, Qwen, conducted a large-scale “distillation” campaign between April and June 2026. Distillation refers to the practice of using outputs from a more advanced AI model to train another model, potentially allowing competitors to replicate capabilities without incurring the full cost of development.
Anthropic alleges that the campaign generated more than 28.8 million interactions with Claude through nearly 25,000 fraudulent accounts. The company claims the effort was designed to harvest Claude’s reasoning, coding, and advanced AI capabilities to accelerate the development of Alibaba’s own models. Anthropic argues that such activity represents a significant intellectual property and security threat, particularly as global competition in AI intensifies.
The allegations come amid broader concerns in the United States about foreign attempts to gain access to advanced AI technologies. Anthropic has previously reported similar extraction efforts involving other Chinese AI firms, including DeepSeek, Moonshot AI, and MiniMax. The company is urging stronger cooperation between governments and AI developers, along with tougher measures to prevent unauthorized access to frontier AI systems and their underlying capabilities.
Alibaba has not publicly responded in detail to the accusations, and the claims have not been independently verified. Nevertheless, the dispute highlights the growing geopolitical tensions surrounding artificial intelligence, where access to advanced models is increasingly viewed as both a commercial advantage and a matter of national security. As AI becomes a strategic technology, conflicts over intellectual property, model security, and international competition are likely to become more frequent.
