Anthropic accuses Alibaba of 'largest campaign' to steal AI capabilities

US artificial intelligence giant Anthropic has accused Chinese e-commerce firm Alibaba of orchestrating the largest known campaign to illicitly extract the capabilities of its Claude AI model—using thousands of fraudulent accounts to carry out almost 29 million exchanges with the system. In a letter sent to two members of the US Congress on 10 June, the San Francisco-based company told Senators Tim Scott and Elizabeth Warren that operators linked to Alibaba had targeted Claude's most valuable features, including its ability to handle longer, more complex tasks and its decision-making processes. The company described the effort as a series of “distillation attacks”, in which answers from a stronger AI model are extracted to train a weaker one on an “industrial scale”. This allows Chinese companies to harvest and repackage US AI capabilities as their own, Anthropic said, turning “hundreds of billions of dollars in American investment and [research and development] into a massive subsidy for our geopolitical competitors”. The letter also cited other alleged attacks that posed a threat to the US military, referencing Department of Defense claims that Alibaba, along with car maker BYD and tech firm Baidu, are tied to the Chinese military—allegations the companies have denied. Alibaba this week sued the US government in a bid to be removed from the Pentagon’s blacklist. Anthropic urged Congress to penalise companies behind such attacks and to ramp up measures to prevent US technology from being stolen. The accusation echoes previous claims by OpenAI, which has also accused Chinese groups of employing the same practice. Anthropic, a leading AI developer alongside 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 more advanced models, such as Mythos, have raised cybersecurity concerns over their ability to target weaknesses in computer systems.