Anthropic, the company behind the Claude AI assistant, has publicly accused Chinese tech giant Alibaba of running a large-scale "distillation" campaign to siphon off Claude's capabilities and use them to improve its own model, Qwen.
According to coverage from outlets including Newsbytes and TweakTown, Anthropic alleges that Alibaba set up roughly 25,000 fake accounts and used them to query Claude some 28.8 million times, then trained Qwen on the results. In AI, "distillation" refers to using one model's outputs to teach another — effectively letting a rival learn from your system's answers rather than building that intelligence from scratch.
Reports from Ars Technica frame the accusation pointedly, claiming Alibaba "defied Trump" to extract Claude's abilities. Several outlets, including Decrypt, note that Anthropic is urging Congress to crack down on AI distillation by Chinese competitors, suggesting the company wants a policy response, not just a corporate dispute.
The accusation lands amid broader friction between Alibaba and Washington. As International Business Times and others report, Alibaba is simultaneously suing the Pentagon over its inclusion on a U.S. military blacklist — underscoring how tangled the company's relationship with American institutions has become.
It's worth noting what these sources do not establish. The reports relay Anthropic's claims; they do not confirm them independently, and they do not include a detailed response from Alibaba defending or denying the specifics. As AI Magazine's framing — posed as a question, "Did Alibaba Illegally Extract Anthropic's AI Capabilities?" — makes clear, the legal and factual questions remain open.
Why it matters: the dispute spotlights one of the central anxieties in the AI race — that a leader's expensive breakthroughs can be cheaply copied by querying the finished product, turning a technical practice into a geopolitical flashpoint between U.S. and Chinese tech.