Anthropic Combats Illicit Distillation Practices Involving Chinese Competitors
Anthropic has raised concerns about foreign competitors using illegal distillation methods to replicate its Claude models, particularly from China. The company's head of threat intelligence, Jacob Klein, described an "illicit ecosystem" on the dark web that enables unauthorized access and training of competing AI systems at a lower cost. The issue is intensifying as Anthropic prepares for a potential IPO in October with a private valuation near $1 trillion. Chinese startups like Moonshot AI have been accused by Anthropic of distilling its models to create cheaper alternatives, such as the Kimi K3 model that has gained traction globally. The U.S. government also weighed in on the issue, stating in an April memo that distillation undermining American research is "unacceptable" and pledging action against foreign actors. Distillation remains a contentious topic within AI development circles, with some advocating for minimal regulation to foster competition while others push for stronger IP protections.
Topics: AI competition dark web distillation theft