Advertisement
UK

Anthropic accuses Alibaba of ‘largest’ illicit extraction of AI capabilities

Anthropic accuses Alibaba of 29 million illicit exchanges to extract capabilities from its Claude AI model.

UK

Anthropic accuses Alibaba of ‘largest’ illicit extraction of AI capabilities

Almost 29 million times, operators linked to Chinese e-commerce giant Alibaba interrogated Anthropic's Claude AI using thousands of fraudulent accounts – what the US artificial intelligence company has called the largest campaign to illicitly extract its capabilities.

In a letter sent to US Senators Tim Scott and Elizabeth Warren on 10 June, San Francisco-based Anthropic accused Alibaba of “brazenly” and “illicitly” extracting Claude’s most valuable features, including its ability to handle longer tasks and its decision-making approach. The technique, known as a “distillation attack,” extracts answers from a stronger AI model to train a weaker one, Anthropic said.

Anthropic accuses Alibaba of 29 million illicit exchanges to extract capabilities from its Claude AI model.

The company urged Congress to penalise the companies behind such attacks and ramp up measures to prevent US technology from being stolen. “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.

Advertisement

Anthropic’s letter also cited other alleged attacks it said posed a threat to the US military, referencing claims by the US Department of Defense that Alibaba, car maker BYD and tech firm Baidu are tied to the Chinese military. The companies have denied the allegations, and Alibaba this week sued the US government to have its name removed from the Pentagon blacklist.

US developers have previously accused Chinese competitors of using distillation attacks to train models rivalling American AI at a fraction of the cost. OpenAI has also accused Chinese groups of employing the same practice.

Anthropic, a leading AI developer alongside ChatGPT-maker OpenAI, is gearing up for a blockbuster stock market debut that could make it one of the most valuable companies in the world. But some of its advanced models, such as Mythos, have raised cybersecurity concerns over their ability to target weaknesses in computer systems.

Advertisement

The BBC has contacted Alibaba for comment.

Advertisement
Advertisement