Anthropic accuses Alibaba of campaign to 'brazenly' and 'illicitly' extract AI capabilities

In a recent development, Anthropic has formally accused Alibaba of conducting a significant campaign to illicitly extract its artificial intelligence technology. This allegation was outlined in a letter sent to the U.S. Senate Committee on Banking, Housing, and Urban Affairs, which was confirmed by CNBC on Wednesday. The correspondence, addressed to Senators Tim Scott (R-S.C.) and Elizabeth Warren (D-Mass.), details an incident described as the "largest known distillation attack" on Anthropic to date. Distillation is an AI training technique that involves creating a simpler model based on the outputs of a more advanced one. According to the letter, individuals linked to Alibaba and its AI lab executed an astounding 28.8 million interactions with Anthropic's models through approximately 25,000 fraudulent accounts between April 22 and June 5. An Anthropic spokesperson emphasized the necessity of a united front between government and industry to combat such threats, stating, "We believe combating the threat of illicit distillation requires coordinated action, and we will continue working with Congress and the Administration to maintain American AI leadership." Meanwhile, Alibaba has not yet responded to requests for comment from CNBC. This letter comes on the heels of a memorandum issued by the White House Office of Science and Technology Policy, which aimed to assist AI companies in detecting and countering large-scale distillation efforts. Anthropic noted that Alibaba proceeded with its alleged attacks despite previous warnings from the Trump Administration. Earlier this year, Anthropic had identified three other significant distillation campaigns from different AI labs, highlighting an alarming trend of increasing sophistication and intensity in such attacks. The company encouraged collaboration among AI industry players, cloud service providers, and policymakers to address these challenges. However, Anthropic's recent engagements with policymakers have faced hurdles. Earlier this month, the company received an export control directive mandating that access to its latest AI models, Claude Fable 5 and Mythos 5, be suspended for all foreign nationals, including its own foreign employees. While the government cited concerns related to national security, it did not elaborate on the specifics. Anthropic's senior staff traveled to Washington, D.C., for discussions with the Trump administration, and the company indicated that both parties are working diligently to resolve the situation, though a timeline for restoring access to the models remains unclear.