Anthropic: Alibaba Ran Largest Claude Extraction
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic alleged in a 10 June letter to Senators Tim Scott and Elizabeth Warren that Alibaba-linked operators ran 'the largest campaign to illicitly extract Claude's capabilities,' carrying out nearly 29 million exchanges with Claude via thousands of fraudulent accounts
- Anthropic said the campaign relied on 'distillation attacks' — extracting answers from a stronger model to train a weaker one — and specifically targeted Claude's handling of longer, complex tasks and its decision-making approach
- Anthropic urged Congress to penalize companies behind such attacks, framing them as turning 'hundreds of billions of dollars in American investment' into 'a massive subsidy for our geopolitical competitors'
- The letter also cited other alleged attacks threatening the US military, referencing Pentagon claims that Alibaba, BYD, and Baidu are tied to the Chinese military — allegations all three companies have denied
- Alibaba sued the US government this week to get removed from the Pentagon blacklist, and the BBC has contacted Alibaba for comment on the distillation allegations
- OpenAI has previously accused Chinese groups of employing the same distillation tactic to train rival models at a fraction of the cost
- Anthropic is preparing for a blockbuster stock market debut alongside OpenAI, even as some of its advanced models like Mythos have drawn cybersecurity concerns
Why it matters: Anthropic's 29 million-exchange figure escalates a US-China AI rivalry that now has a concrete dollar-scale framing: the company argues distillation attacks effectively transfer hundreds of billions in US R&D to competitors. With Alibaba already suing to escape the Pentagon blacklist, the accusation lands amid an active legal fight, and both Anthropic and OpenAI are heading toward market debuts where regulatory and geopolitical narratives will shape valuations.
Ask SkimNews
