Anthropic Accuses Alibaba of Large-Scale AI Model Extraction in Letter to US Congress
Artificial intelligence company Anthropic has accused Chinese technology giant Alibaba of carrying out what it describes as the largest known attempt to extract the capabilities of its Claude AI model through fraudulent means.
In a letter dated June 10 and addressed to US Senators Tim Scott and Elizabeth Warren, the San Francisco-based company alleged that operators linked to Alibaba conducted nearly 29 million interactions with Claude using thousands of fake accounts. Anthropic described the operation as “the largest campaign to illicitly extract Claude’s capabilities.”
According to the company, the activity involved so-called “distillation attacks” – a technique in which responses generated by a more advanced AI model are collected and used to train another model with similar capabilities.
Anthropic claimed the operators focused on Claude’s most advanced features, including its ability to handle complex, long-form tasks and its decision-making processes.
The company argued that such attacks are being conducted on an “industrial scale,” enabling Chinese firms to replicate and commercialise American AI innovations.
In the letter, Anthropic urged US lawmakers to strengthen legal penalties against organisations involved in AI model extraction and introduce tougher safeguards to prevent the theft of American artificial intelligence technology.
“Distillation attacks turn hundreds of billions of dollars in American investment and [research and development] into a massive subsidy for our geopolitical competitors,” Anthropic wrote.
The company also argued that such practices pose broader national security concerns, referencing claims by the US Department of Defense that Alibaba, along with companies including BYD and Baidu, has links to the Chinese military.
Alibaba, BYD and Baidu have consistently denied allegations of military affiliation. Earlier this week, Alibaba filed a lawsuit against the US government seeking to remove its name from a Pentagon blacklist.
The BBC said it had contacted Alibaba for comment on Anthropic’s allegations and had also requested additional details from Anthropic regarding the claims.
The dispute adds to growing tensions between American and Chinese AI developers over the use of model distillation. Several US companies have previously alleged that Chinese competitors have used similar techniques to build competing AI systems at significantly lower development costs. OpenAI has also publicly accused Chinese groups of employing comparable methods.
Anthropic, one of the leading developers of generative AI alongside OpenAI, is widely expected to pursue a public listing in the future, with analysts predicting it could become one of the world’s most valuable technology companies.
At the same time, some of Anthropic’s advanced AI models, including Mythos, have drawn scrutiny from cybersecurity experts over concerns about their ability to identify vulnerabilities in computer systems.
