Anthropic Accuses Alibaba of Large-Scale AI Model Theft
Distillation Attacks Target Claude's Advanced Capabilities
US artificial intelligence company Anthropic has accused Chinese e-commerce and technology firm Alibaba of extracting capabilities from its Claude AI model through distillation attacks.
Anthropic described this as the largest known campaign to illicitly extract Claude's capabilities.
In a letter sent to US Senators Tim Scott and Elizabeth Warren on June 10, Anthropic alleged that operators linked to Alibaba carried out nearly 29 million exchanges with Claude using thousands of fraudulent accounts.
What Are Distillation Attacks?
Distillation attacks involve extracting answers from a stronger AI model to train a weaker one. Anthropic claims these attacks are conducted on an industrial scale by Chinese companies to repurpose US AI capabilities.
The alleged attacks specifically targeted Claude's advanced features, including:
- Its ability to handle longer, complex tasks
- Its sophisticated decision-making approach
Broader Context and Allegations
The letter also cited other alleged attacks that Anthropic says pose a threat to the US military. The company referenced US Department of Defense claims linking Alibaba and other firms—such as BYD and Baidu—to the Chinese military; those companies have denied such ties.
Alibaba has sued the US government to be removed from a Pentagon blacklist.
Other US developers, including OpenAI, have previously accused Chinese groups of using similar distillation methods.
Response and Next Steps
The BBC has contacted Alibaba for comment and requested more details from Anthropic.
Anthropic urged Congress to penalize companies behind such attacks and strengthen measures against theft of US technology.