Anthropic has accused Chinese e-commerce giant Alibaba of orchestrating a massive campaign to illicitly harvest capabilities from its Claude models, according to a June 10 letter sent to U.S. lawmakers.
The letter, addressed to Sens. Tim Scott (R-S.C.) and Elizabeth Warren (D-Mass.) of the Senate Banking Committee, claims that operators affiliated with Alibaba and its artificial intelligence (AI) lab, Alibaba Qwen, executed the largest known “distillation attack” in Anthropic’s history. Between April 22 and June 5, attackers allegedly deployed roughly 25,000 fraudulent accounts to conduct nearly 29 million exchanges with Claude.
AI distillation is a training methodology wherein a smaller, less capable model is trained on the outputs of a more powerful system. According to Anthropic, this practice allows foreign competitors to bypass hundreds of billions of dollars in American research and development, effectively turning U.S. investments into a subsidy for geopolitical rivals. The firm alleged that Alibaba targeted Claude’s complex decision-making and long-context processing to accelerate China’s development toward rivaling Anthropic’s advanced Mythos Preview capabilities.
“We believe combating the threat of illicit distillation requires coordinated action between government and industry,” an Anthropic spokesperson said. He urged Congress to penalize participating firms and bolster intellectual property defenses.
The escalating friction arrives amid a broader crackdown by Washington, D.C., on foreign access to American technology. In April, the White House accused Beijing of stealing U.S. intellectual property on an “industrial scale.” Anthropic itself previously flagged similar distillation campaigns in February conducted by other prominent Chinese AI entities, including DeepSeek, Moonshot, and MiniMax.
Geopolitical tensions are further compounded by recent regulatory maneuvers. Alibaba was recently added to the Pentagon’s list of Chinese military-linked companies, a designation the e-commerce giant is currently contesting in federal court. Meanwhile, U.S. authorities are walking a tightrope; the Commerce Department recently held off on trade-blacklisting DeepSeek to avoid further escalating tensions with Beijing.
However, Anthropic’s aggressive engagement with Washington has brought its own operational hurdles.
On June 12, just two days after the letter was dispatched, the Commerce Department imposed strict export controls on Anthropic’s cutting-edge Mythos and Fable models over concerns they could be exploited by foreign military intelligence.
The directive forced Anthropic to temporarily suspend global access to these systems, affecting even its own foreign national employees. Anthropic confirmed senior executives have met with Trump administration officials to resolve the restrictions, though the advanced models remain offline.
Alibaba did not immediately respond to requests for comment regarding the allegations. The incident highlights the growing urgency among U.S. technology leaders and policymakers to secure proprietary AI models, which are increasingly viewed as vital to national security and global economic dominance.
