Using a technique it calls a "distillation attack," AI giant Anthropic has accused Chinese tech firm Alibaba of illicitly siphoning capabilities from its Claude model in what it describes as the largest known theft of its kind against the company. The bombshell accusation was detailed in a letter sent to two members of the US Congress.
Anthropic alleges the campaign's goal was to steal the AI's most valuable features, including its complex task-handling and unique decision-making processes. Operators linked to Alibaba reportedly conducted almost 29 million exchanges with Claude using thousands of fraudulent accounts.
The distillation technique involves a competitor systematically prompting a more advanced AI to harvest its reasoning patterns. These stolen responses are then used to train a weaker, internal AI, allowing a company to bypass millions of dollars in research and development costs.
Theft to Bypass Billions in R&D
In the letter, dated June 10 and sent to US Senators Tim Scott and Elizabeth Warren, Anthropic condemned the effort as a "brazen" and "illicit" campaign. The company claims these attacks are carried out on an "industrial scale" to allow Chinese firms to repackage US AI capabilities as their own.
"These distillation attacks are carried out illicitly, systematically, and at an industrial scale to harvest US AI capabilities across frontier labs and repackage them as their own," the letter stated.
This incident echoes a White House accusation from April that China was stealing intellectual property from U.S. AI labs. In its letter, Anthropic supported U.S. government efforts to combat these threats through intelligence sharing.
Call for Government Action
Anthropic urged Congress to penalize the companies behind such attacks and implement stronger measures to prevent the theft of American technology. The firm argues these actions pose a significant threat to national security and the U.S. tech industry's competitive edge.
"Distillation attacks turn hundreds of billions of dollars in American investment and [research and development] into a massive subsidy for our geopolitical competitors," the company wrote.
The accusation puts a spotlight on a wider pattern of behavior. In February, Anthropic identified a similar campaign by Chinese AI startup DeepSeek and two other labs.
Alibaba, for its part, was added to the Pentagon's list of Chinese military companies earlier this month—a designation the company is currently challenging. The BBC has reached out to Alibaba for comment on Anthropic's latest claims.