Anthropic Accuses Alibaba of Illegally Accessing AI Technology

Anthropic has charged the Chinese technology company Alibaba with “brazen” and “illegal” efforts to siphon off its artificial intelligence capabilities, claiming this is “the largest distillation attack ever known against Anthropic.”

In a letter to the Senate Banking, Housing and Urban Affairs Committee earlier this month, Anthropic outlined these alleged violations of code, as reported by Bloomberg.

The technique at the center of this dispute, known as distillation, involves utilizing the output from more advanced AI models to create smaller, less capable versions. 

Led by Dario Amodei, Anthropic stated that individuals associated with Alibaba’s Kwen AI Lab engaged in 28.8 million interactions with its Claude model using around 25,000 fake accounts between late April and early June, per Bloomberg’s findings.

“We think that addressing the threat of illegal distillation demands a coordinated effort between government and industry, and we will keep collaborating with Congress and the administration to preserve America’s AI leadership,” an Anthropic spokesperson expressed in a statement.

There hasn’t been an immediate response from Alibaba regarding this claim.

The letter from Anthropic, addressed to Senators Tim Scott (R-SC) and Elizabeth Warren (D-MA), comes less than two months after a memorandum from the White House Office of Science and Technology Policy aimed at assisting AI companies in recognizing and adjusting to mass distillation.

Anthropic accused Alibaba of disregarding warnings from the Trump administration.

Experts have raised alarms over a surge of powerful, cost-effective AI models from China that could endanger America’s competitive edge in this field.

One notable recent release is GLM-5.2 from China’s z.AI, launched on June 16 and focused on coding projects. This model is claimed to be nearly as advanced as those from Anthropic, OpenAI, and Google.

In February, Anthropic mentioned it had detected three large-scale distillation efforts from other Chinese institutes: DeepSeek, Moonshot, and MiniMax.

Anthropic suggested that the sophistication and intensity of these campaigns are increasing, urging policymakers and tech companies, including cloud providers, to join forces in counteracting potential future attacks.

Recently, Anthropic has faced several challenges from the White House, particularly regarding concerns surrounding its powerful “Fable” and “Myth” models, putting the company in a precarious position to address these security issues.