The number is 29 million. That is the figure at the heart of a formal accusation now before the US Senate. Nearly 29 million interactions.
Thousands of fraudulent accounts. All aimed at one target: the Claude AI model built by San Francisco-based Anthropic.
Those interactions were not random queries. Anthropic says they were part of a coordinated extraction campaign, and it has named Alibaba as the orchestrator. The accusation landed in a letter dated June 10, sent to Senators Tim Scott and Elizabeth Warren.
In it, the company describes the effort as the largest extraction campaign of its kind. The technique is called a distillation attack.
It works like this: an operator feeds prompts to a powerful, existing AI model — in this case, Claude. The model generates responses. Those responses are collected, stored, and then used to train a less capable model.
It is a shortcut. Instead of building an advanced AI from scratch, you harvest the intelligence of one that already exists. Anthropic claims the operators focused on Claude’s most valuable functions.
The ability to handle long and complex tasks was targeted. So were the model’s methods of reasoning and decision-making.
These are not trivial features. They are the core capabilities that make a high-end AI model useful and expensive to develop. The letter argues that such attacks are being conducted on an “industrial scale.” The phrase matters.
It suggests this was not a one-off experiment by a few engineers. It implies a sustained, organized operation.
Anthropic’s argument is that Chinese companies are using this method to harvest advanced US AI technologies and repackage those capabilities into their own systems. This accusation did not emerge in a vacuum. The technology relationship between the United States and China has been defined by mutual suspicion for years.
Each side has accused the other of theft, espionage, and unfair practices. This letter adds a specific, detailed charge to that long history. Anthropic also cited other alleged attacks.
Those, the company said, posed a threat to the US military. The letter did not provide further details on those attacks, but the implication is clear: extraction campaigns are not limited to commercial advantage.
They have national security dimensions. The company is now urging Congress to impose penalties on organizations involved in such activities. It wants a legal and economic response.
Whether that happens depends on the senators who received the letter and on the broader legislative body. For Alibaba, the accusation is a serious one.
The company has not yet issued a public response to the specific claims in the June 10 letter. But the charge is on the record, in writing, sent to two influential members of the US Senate. Twenty-nine million interactions.
That is a lot of data. If Anthropic’s account is accurate, it represents a methodical effort to strip-mine a piece of American AI technology. The company is betting that the scale of the operation will make it impossible for Congress to ignore.
The letter is dated June 10. It was reported by state and international outlets citing the BBC.
The numbers are specific. The accusation is direct. The stakes are high.
