Anthropic has accused Alibaba of orchestrating a broad effort to gain unauthorized access to its Claude artificial intelligence model, in what the US startup described as the largest known attempt by a Chinese company to copy the work of a leading American AI lab.

In a letter sent to several US senators and White House officials, Anthropic said operators linked to Alibaba's Qwen AI lab used thousands of fraudulent accounts to probe Claude's capabilities. The company said the activity appeared aimed at the model's most valuable functions, including software engineering and agentic reasoning.

The allegation highlights growing tension over access to advanced AI systems and the potential for so-called distillation, a process in which one model is trained to imitate the behavior of another. Anthropic has said it limits access to its products in China, and the company argued that the campaign it identified was intended to get around those restrictions.

The startup did not provide details in the material available about how it determined the links to Alibaba, and the report did not include a response from Alibaba. Anthropic's claim nevertheless adds another layer to an increasingly competitive global race to build and control frontier AI models, where companies are guarding both technical capabilities and market access.

Claude is one of Anthropic's flagship products and has been positioned as a model with strong performance in coding and multi-step reasoning tasks. Those are among the same capabilities that appear to have been targeted in the alleged access effort, according to the letter described in the report.

The accusation also underscores the extent to which leading AI labs are worried about their systems being used as templates for rivals. If model outputs can be harvested at scale through fake accounts or other methods, developers say that can help competitors replicate hard-won capabilities without bearing the same research and training costs.

Anthropic's warning was directed at senior US policymakers, suggesting the company wants the issue treated as more than a commercial dispute. The outreach reflects broader concerns in Washington about foreign access to advanced American AI systems, especially when those systems are viewed as strategically important.

Alibaba is one of China's biggest technology companies and has built its own family of AI models through Qwen. Anthropic's claim places the company at the center of one of the most serious public accusations yet involving alleged unauthorized access to a major US model.

The episode comes as AI developers are tightening controls around their systems while trying to expand their reach. Companies are balancing the desire to make powerful models available to customers with fears that those same models can be reverse-engineered, copied, or otherwise exploited by competitors.

Anthropic's allegation, if substantiated, would mark a significant escalation in the debate over how AI models should be protected and who gets to benefit from them. For now, it remains a claim from one of the industry's most prominent startups, aimed at both a rival tech giant and the US government.