Anthropic says Alibaba used fake accounts to access Claude

In short: Anthropic says Alibaba created fake accounts to access Anthropic’s Claude chatbot and collect large amounts of its responses.

What happened

Anthropic, a US company behind the Claude chatbot, accused China’s ecommerce group Alibaba of getting what it called “illicit” access to Claude. Anthropic says it does not offer Claude to Chinese groups, but that Alibaba still found a way in by creating fake user accounts.

In a letter to the US Congress dated June 10, Anthropic said Alibaba used about 25,000 fraudulent accounts. Anthropic said those accounts generated more than 28.8 million “exchanges” with Claude, meaning back and forth messages between a user and the chatbot.

Anthropic said the effort aimed to pull out some of Claude’s most valuable skills, including “software engineering” and longer, multi step tasks. The company also pointed to “distillation,” which is a common method for making a smaller AI by learning from a bigger one (like copying a chef’s cooking style by tasting many dishes). Anthropic argues it becomes a problem when it breaks rules or is used to copy advanced capabilities.

Alibaba did not respond to a request for comment on Anthropic’s claims, according to the report. Alibaba has denied having any connection to China’s military and is separately asking a US court to remove it from a Pentagon list of companies with alleged military ties.

Why it matters

This dispute shows how tense the AI race has become between US and Chinese companies and governments. For regular people, it could affect what AI tools are available in different countries, how tightly companies lock down their products, and whether lawmakers add new penalties for companies accused of copying AI systems.

Source: Financial Times