**Anthropic Accuses Alibaba of Illicitly Extracting AI Capabilities**
In a significant development in the artificial intelligence sector, US-based AI firm Anthropic has leveled serious allegations against Chinese technology giant Alibaba. The San Francisco-based company claims that Alibaba has engaged in a "brazen" and "illicit" campaign to extract capabilities from its Claude AI model.
According to a letter addressed to US Senators Tim Scott and Elizabeth Warren, which was obtained by the BBC, Anthropic alleges that operators associated with Alibaba executed nearly 29 million exchanges with the Claude model using thousands of fraudulent accounts. The company described this operation as the largest extraction campaign of its kind, raising concerns about the security of US technological advancements.
Anthropic's letter, dated June 10, outlines that the extraction was conducted through a method known as "distillation attacks." This technique involves extracting responses from a more advanced AI model—like Claude—to train a less sophisticated one. The company claims that Alibaba-linked operators specifically targeted Claude's most valuable features, including its capacity to handle complex tasks and its decision-making processes.
The allegations have sparked a broader conversation about the potential risks posed by foreign entities to US technological innovations. Anthropic has urged Congress to take action against companies engaging in such practices and to enhance protective measures against the theft of American technology.
In its correspondence, Anthropic highlighted that these distillation attacks are conducted on an "industrial scale," allowing Chinese firms to harvest and repackage US AI capabilities for their own use. The company emphasized that such activities turn significant American investments in research and development into a competitive advantage for geopolitical rivals.
The letter also referenced other alleged attacks that Anthropic claims pose a threat to US national security, particularly concerning the military. The company cited assertions from the US Department of Defense, which links Alibaba and other major Chinese companies, including car manufacturer BYD and tech firm Baidu, to the Chinese military. These allegations have been denied by the companies involved, with Alibaba recently filing a lawsuit against the US government to remove its name from a Pentagon blacklist.
Anthropic's claims are not isolated; other US developers have previously accused Chinese competitors of utilizing similar distillation attacks to create models that can compete with American AI technology at significantly lower costs. OpenAI, the creator of ChatGPT, has also reported similar practices among Chinese groups.
As Anthropic prepares for a potential stock market debut, it stands alongside OpenAI as one of the leading AI developers in the world. However, the company has faced scrutiny over some of its advanced models, such as Mythos, which have raised cybersecurity concerns due to their capabilities in identifying vulnerabilities in computer systems.
The ongoing tensions between US and Chinese technology firms highlight the complexities of international competition in the rapidly evolving AI landscape. As both nations strive to establish dominance in AI development, the implications of these allegations could have far-reaching consequences for the industry and national security.
The situation remains dynamic, with Anthropic awaiting a response from Alibaba regarding its claims. As the discourse around AI security and intellectual property intensifies, the need for robust protective measures and international cooperation becomes increasingly critical.