Alibaba Group (NYSE:BABA) stock declined roughly 3% on Wednesday after AI developer Anthropic accused operators associated with Alibaba's Qwen AI lab of illicitly scraping the company's Claude models.
In a letter addressed to U.S. senators and White House officials, Anthropic described what it called a coordinated and large-scale effort to bypass usage restrictions. The startup reported that the campaign ran from April through June, and said it amounted to 28.8 million exchanges with the Claude model executed via almost 25,000 fraudulent accounts.
Anthropic characterized the activity as the largest known attempt by a Chinese company to exploit a leading U.S. AI lab. The company also said the pattern of activity resembled other scraping attempts it has flagged from Chinese developers earlier this year.
The startup warned policymakers that Alibaba and other Chinese AI labs are systematically harvesting data from top-tier U.S. models. Anthropic used the term "adversarial distillation" to describe the practice, noting that it enables foreign competitors to build competing chatbots at a small fraction of the original training expense.
"These attacks are carried out illicitly, systematically, and at industrial scale to harvest US AI capabilities across frontier labs and repackage them as their own without incurring the training and R&D costs," Anthropic wrote in its letter to the Trump administration.
Anthropic emphasized that it explicitly blocks access to its products from within China. The company argued the alleged scraping therefore constitutes a clear breach of its distribution controls and terms of service. In response to the incident, Anthropic urged U.S. authorities to impose stricter guardrails to stop similar campaigns.
Beyond the immediate market reaction for Alibaba shares, Anthropic's account raises questions about intellectual property protection, the robustness of model safety measures, and the potential commercial incentives that drive adversarial scraping at scale. The startup also warned that models reconstructed through such methods often lack essential safety guardrails.
Summary
Anthropic has accused operators linked to Alibaba's Qwen AI lab of conducting a large-scale scraping campaign against its Claude models. The company reported 28.8 million exchanges occurring from April to June across about 25,000 fraudulent accounts, and has asked U.S. policymakers for stronger protections. Alibaba's shares fell about 3% after the allegations surfaced.
- Key points
- Anthropic reported a campaign running April through June involving 28.8 million exchanges and nearly 25,000 fraudulent accounts.
- The startup said the operation mirrors prior scraping attempts by Chinese developers and described the practice as "adversarial distillation."
- Impacted areas include technology and AI markets, with possible repercussions for investor sentiment toward companies linked to alleged scraping.
- Risks and uncertainties
- Regulatory risk - Policymakers may face pressure to introduce stricter oversight of cross-border AI data practices, affecting AI labs and cloud services.
- Intellectual property and legal risk - Allegations of large-scale scraping raise potential legal and contractual disputes between AI developers and alleged infringers.
- Safety risk - Models rebuilt through adversarial distillation may lack critical safety guardrails, presenting potential harms if deployed.
Anthropic's account, as presented in its communications to U.S. officials, lays out a concentrated and resource-intensive campaign to extract model interactions for reuse. The firm has framed the practice as systematic and industrial in scale and has called on U.S. authorities to act to prevent further occurrences.