Anthropic accuses Alibaba of stealing its AI model

Thu Jun 25 2026 07:30:35 GMT+0000 (Coordinated Universal Time)

US AI maker Anthropic claims Chinese e‑commerce behemoth Alibaba ran thousands of fake accounts to siphon its Claude model’s data.

Anthropic sent a letter to US Senators Tim Scott and Elizabeth Warren alleging that Alibaba linked operators conducted a large‑scale distillation attack, pulling Claude AI answers through fraudulently created accounts. The company says the operation exposed advanced Claude features, such as long‑form reasoning, for commercial use by Alibaba and other Chinese firms tied to the military. Anthropic urges Congress to impose penalties to protect US AI research and demands tighter safeguards against such data harvesting. Alibaba denies the allegations and has recently sued the Pentagon for being listed on a blacklist. Industry experts note that similar attacks have been reported against other AI companies, including OpenAI.

Anthropic, the US‑based AI firm behind its Claude model, wrote a formal letter to Senators Tim Scott and Elizabeth Warren accusing Alibaba of illegally extracting Claude’s intelligence through a large scale distillation attack.

In the letter, Anthropic claimed that near 29 million requests were made, using thousands of forged accounts, to harvest Claude’s answers. The operation targeted the model’s key strengths, such as reasoning over long text passages and decision‑making processes, according to the company.

Anthropic urges lawmakers to penalise the companies involved and strengthen measures that prevent American AI tech from being stolen by foreign competitors. The company also cited similar concerns from the US Department of Defense regarding potential military threats.

Alibaba has denied the accusations, and this week it filed a lawsuit against the U.S. government to remove it from a Pentagon blacklist. Other Chinese firms, including Baidu and BYD, have faced similar allegations, some of which also claim ties to the Chinese military.

Industry observers note that AI giants such as OpenAI have previously reported the same distillation‑attack pattern from Chinese competitors, raising broader concerns about AI security and intellectual property protection.