Anthropic Raises Concerns Over Alleged AI Extraction by Alibaba
US artificial intelligence developer Anthropic has accused Chinese e-commerce and technology company Alibaba of "brazenly" and "illegitimately" extracting the capabilities of its Claude AI model. The allegations were detailed in a letter sent by Anthropic to two members of the US Congress.
According to the San Francisco-based company, operators linked to Alibaba engaged in nearly 29 million interactions with Claude, utilizing thousands of fraudulent accounts. Anthropic characterized this as the most extensive extraction campaign of its kind to date. The company has urged Congress to impose penalties on entities involved in such activities and to enhance measures designed to prevent the theft of US technological advancements.
Details of the Alleged Extraction Campaign
The letter from Anthropic, dated June 10 and addressed to US Senators Tim Scott and Elizabeth Warren, specifically accused New York Stock Exchange-listed Alibaba of orchestrating "the largest campaign to illicitly extract Claude's capabilities."
Anthropic claims the campaign was conducted through "distillation attacks." This method involves extracting responses from a more powerful AI model to train a less capable one. The company stated that Alibaba-linked operators specifically targeted Claude's most valuable features, including its proficiency in handling extended and complex tasks, as well as its decision-making processes. Anthropic asserted that these types of attacks are executed on an "industrial scale" to enable Chinese companies to acquire and rebrand US AI capabilities as their own.
Broader Implications and Previous Allegations
The letter also referenced other alleged attacks, which Anthropic suggested could pose a threat to US national security. "Distillation attacks transform hundreds of billions of dollars in American investment and [research and development] into a massive subsidy for our geopolitical competitors," Anthropic stated. The company cited claims from the US Department of Defense linking Alibaba and several other major firms, such as car manufacturer BYD and tech company Baidu, to the Chinese military. These companies have denied such accusations, and Alibaba recently initiated legal action against the US government to be removed from the Pentagon's blacklist.
This is not an isolated incident; US developers have previously accused Chinese competitors of employing distillation attacks to train their AI models to compete with American technology at a significantly reduced cost. OpenAI, another prominent AI developer, has also previously reported similar practices by Chinese groups.
Anthropic, alongside OpenAI, is a leading AI developer and is reportedly preparing for a significant stock market debut, which could position it among the world's most valuable companies. However, some of Anthropic's more advanced models, like Mythos, have raised cybersecurity concerns due to their potential to identify vulnerabilities in computer systems.
Source: Anthropic accuses Chinese rival Alibaba of illicitly extracting AI capabilities