The American artificial intelligence company Anthropic has accused the Chinese technology giant Alibaba of trying to illicitly copy capabilities from its Claude AI models, in a dispute that highlights the intensifying competition between US and Chinese developers of advanced AI. The accusation has not been proven, and Alibaba has not publicly responded.
The accusation
Anthropic set out the allegation in a letter to US lawmakers, CNBC reported. The company said operators it linked to Alibaba had used large numbers of fraudulent accounts to send millions of queries to its Claude models over a period of weeks, in what it characterized as an effort to harvest the systems' capabilities "illicitly, systematically, and at an industrial scale."
Anthropic said the activity focused on Claude's most valuable skills, such as software engineering, and continued despite earlier US government warnings to Chinese AI developers about improperly accessing American AI systems, according to The Next Web. The specific figures cited — including the number of accounts and queries — come from Anthropic's own account and have not been independently verified.
What 'distillation' means
At the center of the claim is a practice known as model distillation. It involves repeatedly querying a more capable AI model, collecting its answers, and using that output to train a cheaper model to imitate the original's behavior. Done with permission, distillation is a routine technique; done against a competitor's system without authorization, it can be a way to approximate years of expensive research at a fraction of the cost. Critics also note that a model trained this way may not inherit the safety safeguards built into the original.
A wider pattern of disputes
The accusation is the latest in a series of similar claims among AI developers. Anthropic has previously said it identified other large-scale extraction efforts it attributed to Chinese AI firms. In early 2025, OpenAI raised comparable concerns about the Chinese startup DeepSeek after its low-cost model rattled the industry, alleging its systems may have been used without authorization — claims that did not lead to formal legal action.
The backdrop is a broader contest between the United States and China over advanced AI, which Washington increasingly treats as a strategic priority. US officials have moved to restrict access to the most capable American models, and some lawmakers have floated penalties for foreign firms found to have improperly extracted US AI model outputs. Alibaba is separately contesting a US Defense Department designation it has called baseless.
Alibaba's silence
As of publication, Alibaba had not responded publicly to Anthropic's allegations, and requests for comment from several news organizations went unanswered. A lack of response is not an admission; companies routinely decline to comment on accusations while weighing legal or diplomatic options.
It is not clear whether Anthropic intends to pursue litigation or whether US authorities will act on its disclosures. The letter to Congress appears aimed at building support for a coordinated government and industry response. The episode underscores a growing tension in the AI industry: as leading models become more powerful, so do the incentives — and the techniques — to extract their capabilities, while the legal and technical means to prevent it remain underdeveloped.
