US artificial intelligence company Anthropic has accused Chinese technology giant Alibaba of carrying out a large-scale effort to extract capabilities from its Claude AI model. The allegation appeared in a letter reviewed by Reuters and sent to senior members of the US Senate Banking Committee ahead of a scheduled hearing on artificial intelligence.
Anthropic described the activity as a “distillation” campaign, a process in which a less advanced AI system is trained using outputs from a more powerful model. The company said such activity can reduce the performance gap between competing AI systems by copying behavioural patterns and responses.
In its letter, Anthropic said the alleged campaign ran between April 22 and June 5, 2026, and involved more than 28.8 million interactions with Claude, reported Reuters. The company said nearly 25,000 fraudulent accounts took part in generating these exchanges.
Anthropic linked the activity to operators it said were associated with Alibaba and its AI division, Alibaba Qwen.
Anthropic’s claims
Anthropic said the activity aimed to extract advanced capabilities from its Claude model, which it uses in enterprise and consumer AI products. The company said the scale of the operation made it one of the largest known cases of attempted model extraction it had observed.
The letter stated the campaign appeared designed to speed up the development of competing AI systems in China by replicating the performance of Anthropic’s more advanced models, reported Reuters. The company said this could reduce the time needed for rival systems to reach similar capabilities.
Anthropic said the method involved repeated queries to Claude through large numbers of accounts, generating structured outputs that could be used to train other models. It said this type of activity created risks for AI companies by weakening the protection of proprietary systems.
A spokesperson for Anthropic told Reuters that the company supported government efforts to address such risks. “We support coordinated action between industry and policymakers to address emerging threats to advanced AI systems,” the company said.
Why has AI model distillation become a concern?
Anthropic said model distillation has become a growing concern across the AI industry. The company said similar activity had been observed in past cases involving other Chinese AI firms.
In earlier disclosures, Anthropic said Chinese companies such as DeepSeek, Moonshot AI and MiniMax had also attempted similar extraction efforts at different scales, reported Reuters. The company said these activities ranged from hundreds of thousands to millions of interactions with its systems.
Anthropic said these patterns showed increasing attempts to replicate frontier AI models through indirect access rather than direct development. It said the trend raised concerns about intellectual property protection and competition in the global AI sector.
The issue has also drawn attention from US policymakers. The White House earlier said China had engaged in large-scale efforts to access advanced AI technology developed in the United States. Alibaba was also added to a US defence-related list of Chinese companies this month, a designation the company is challenging.
At the same time, US regulators have taken steps that affect AI companies on both sides. The Commerce Department recently imposed restrictions on Anthropic’s advanced models over concerns about possible use in sensitive environments, which led to global access limitations.