Anthropic has accused Alibaba of running a large-scale campaign to probe and potentially copy capabilities from its artificial intelligence systems, according to a letter the U.S. company sent to Senate banking leaders.
The letter, dated June 10 and reviewed by CNBC, was addressed to Senate Banking Committee Chair Tim Scott of South Carolina and ranking member Elizabeth Warren of Massachusetts. In it, Anthropic said Alibaba and operators linked to the company’s AI lab carried out what it described as the largest known distillation attack it has seen against its models.
Distillation is a common AI training technique in which a smaller model learns from the outputs of a larger, more advanced one. Anthropic alleges that the method was used here in a way that crossed into unauthorized extraction of its capabilities.
According to the company, the activity involved 28.8 million exchanges with Anthropic models and around 25,000 fraudulent accounts between April 22 and June 5. Anthropic said the accounts were tied to operators associated with Alibaba and its research lab.
In a statement, an Anthropic spokesperson said the company believes fighting illicit distillation will require coordination between government and industry. The spokesperson added that Anthropic will keep working with Congress and the administration to support U.S. leadership in artificial intelligence.
Alibaba did not immediately respond to CNBC’s request for comment.
Bloomberg first reported the existence of the letter.
The dispute arrives as U.S. policymakers are increasingly focused on how AI companies protect their systems from imitation and misuse. Anthropic said the letter came two months after the White House Office of Science and Technology Policy issued a memorandum that promised support for AI firms seeking to detect and respond to industrial-scale distillation attempts. In its letter, Anthropic said Alibaba had ignored warnings from the Trump administration.
Anthropic has previously warned about similar activity from other Chinese AI labs. In February, the company said it had identified industrial-scale distillation efforts from DeepSeek, Moonshot and MiniMax. At the time, Anthropic said those campaigns were becoming more advanced and urged closer cooperation among AI companies, cloud providers and policymakers.
The company’s latest complaint also comes during a difficult period in its relationship with the U.S. government. Earlier this month, Anthropic said it received an export control directive from the Trump administration ordering it to cut off access to its newest Claude models, Fable 5 and Mythos 5, for any foreign national, including employees of the company.
Anthropic said the government cited national security authorities but did not explain the specific concern. Senior staff then traveled to Washington to meet with administration officials, and the company later said both sides were working quickly to resolve the matter.
Taken together, the developments underscore the rising policy pressure around advanced AI systems and the growing concern over how proprietary model behavior can be copied, studied or repurposed by rivals. For Anthropic, the allegations against Alibaba add another front to a broader debate over AI competition, security and international access to frontier models.