In a letter to US senators Tim Scott and Elizabeth Warren, dated 10 June, the San Francisco-based AI firm alleged that operators tied to the Chinese ecommerce and technology group carried out almost 29m exchanges with Claude using thousands of fraudulent accounts.

Anthropic claimed the activity was designed to extract the model’s most valuable capabilities, including its ability to handle longer and more complex tasks, as well as its approach to reasoning and decision-making.

The company described the alleged activity as a series of “distillation attacks”, where answers from a stronger AI model are used to train a weaker one.

Anthropic told Congress that such attacks were being carried out on an “industrial scale”, allowing Chinese companies to harvest and repackage US-developed AI capabilities at a fraction of the cost of building them independently.

The business urged lawmakers to penalise companies behind alleged attacks of this nature and to strengthen measures aimed at preventing US technology from being copied or transferred to geopolitical rivals.

“Distillation attacks turn hundreds of billions of dollars in American investment and research and development into a massive subsidy for our geopolitical competitors,” Anthropic said in the letter.

The claims come amid rising tension between Washington and Beijing over the development and control of advanced AI technologies.

Anthropic also cited US Department of Defense allegations that Alibaba, alongside firms including BYD and Baidu, has ties to the Chinese military. The companies have denied such claims, while Alibaba has reportedly sued the US government in an attempt to have its name removed from the Pentagon blacklist.

The allegations follow similar claims from other US AI developers, including OpenAI, which has previously accused Chinese groups of using distillation techniques to train rival models.

Anthropic, one of the best-known AI developers alongside OpenAI, is widely expected to pursue a major stock market listing. However, the company’s rapid development of more advanced systems has also prompted concerns in some quarters over cybersecurity and the potential misuse of powerful AI tools.