Anthropic slams Alibaba over alleged AI data theft
Anthropic, the San Francisco‑based AI studio that built the Claude model, publicly named Chinese e‑commerce giant Alibaba in a letter to US senators. The startup claims Alibaba used the so‑called “distillation attacks” – a net of thousands of fake accounts – to scrape Claude’s answers and train its own models, amounting to almost 29 million questionable exchanges. That, the company says, turns U.S. research dollars into a free‑ride for competitors on a global scale.
The letter urges Congress to impose penalties on tech firms that practice such attacks and to strengthen safeguards against future theft. Alibaba, meanwhile, has denied all allegations; the firm is also fighting a lawsuit to be removed from the Pentagon’s blacklist. These disputes highlight a growing friction over AI intellectual property as China’s AI sector seeks to leapfrog the U.S. on a technology battlefield that extends into military realms.
Anthropic’s rise is set against its own ambitious plans for a public listing that could make it one of the world’s most valuable companies, and its advanced models like Mythos are arousing concerns about potential cybersecurity weaknesses. For now, the drama between the two giants promises to push regulators, investors and developers to rethink how the most powerful AI tools are protected and shared across borders.
