Skip to content

Anthropic Accuses Alibaba of Distilling Claude Model

To the point: Anthropic accuses Alibaba of systematically copying Claude through distillation and calls on the US government to impose stricter regulation of Chinese AI companies and export restrictions.

Anthropic is accusing Chinese conglomerate Alibaba of systematically querying its AI software Claude through approximately 25,000 fraudulent accounts between April and June in order to replicate its capabilities through distillation. The company is calling on the US Congress to do more to prevent such attacks and further regulate Chinese companies’ access to AI chips.

According to its own documentation, Anthropic found that actors affiliated with Alibaba made nearly 29 million queries to Claude between late April and early June. In doing so, they used a technique known as distillation: they trained their own AI model to mimic the behavior of Claude’s outputs. Such access violates the platform’s terms of service.

Anthropic views this as a targeted economic espionage attempt and wants American AI companies to be able to warn each other about such attacks in the future without legal concerns. In addition, the company is calling for stricter regulatory measures against Chinese companies that employ distillation, as well as further export restrictions on advanced AI hardware to China. Alibaba has not yet commented on the allegations.

The accusation fits into Anthropic’s longer-term campaign against technology transfer. The company has repeatedly accused Chinese actors of using this method. Elon Musk admitted in court in 2024 that he had trained his AI xAI with OpenAI models in this manner — an approach that Anthropic has criticized as problematic.

Meanwhile, Anthropic itself is at odds with the US government: the Biden administration forced the company to temporarily withdraw its new “Fable 5” model from the market. The official reason cited were alleged security vulnerabilities that could allow the model’s safeguards to be circumvented — in particular a blocking of vulnerability-scanning capabilities. Anthropic disputes the severity of the issues. Negotiations with the government are still ongoing.


Source: www.it-daily.net · Published June 25, 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.1.

Share on: