Anthropic Accuses Alibaba of Large-Scale AI Theft

Thu Jun 25 2026 08:51:22 GMT+0000 (Coordinated Universal Time)

AI firm Anthropic claims Alibaba used illicit methods to harvest its Claude model for competitive advantage, calling the move a massive technology theft.

US AI developer Anthropic alleges that Alibaba carried out a covert operation that extracted the abilities of its Claude model through a technique known as distillation attacks. The letter to Senators Tim Scott and Elizabeth Warren states that Alibaba ran almost 29 million exchanges with Claude using fraudulent accounts to harvest key capabilities. Anthropic warns that such practices drain American investment and provide a boost to China’s AI arms, citing threats to national security. Alibaba and other Chinese firms deny the accusations and Alibaba is already fighting its placement on the US defense blacklist.

Anthropic Accuses Alibaba of Large‑Scale AI Theft

US artificial intelligence company Anthropic has publicly accused Chinese e‑commerce giant Alibaba of extracting key capabilities from its Claude AI model in a “brazen” and “illicit” campaign. In a letter sent to Senators Tim Scott and Elizabeth Warren on 10 June, Anthropic claimed that Alibaba‑linked operators carried out almost 29 million exchanges with Claude using thousands of fake accounts, constituting the largest extraction effort known to the firm.

The company explains the attacks as a type of distillation that pulls information from a larger, stronger model to train a smaller one. Anthropic identified that the campaign focused on Claude’s strengths such as handling long, complex queries and its decision‑making techniques, which would allow competitors to build models that appear equivalent while avoiding the cost of initial research.

Anthropic warned that the technique effectively turns an billions‑of‑dollars investment into a subsidy for China’s AI ambitions, saying the operations pose a risk to U.S. military systems. The letter highlighted additional alleged attacks from other firms such as car manufacturer BYD and tech company Baidu, all of whom reportedly have ties to China’s armed forces.

Alibaba has denied the claims and is currently suing the U.S. government over its removal from the Pentagon blacklist. Chinese competitors have repeatedly been accused of using distillation attacks to cheaply replicate U.S. AI capabilities; OpenAI has also raised similar concerns.

The incident comes as Anthropic prepares for a potential public listing that could place it among the world’s most valuable companies, while its Mythos model has attracted scrutiny for its capacity to identify and exploit software vulnerabilities.

By Osmond Chia, Business reporter