Anthropic Accuses Alibaba’s Qwen Lab of Using 25,000 Fake Accounts to Copy Claude

  • Anthropic says nearly 25,000 fraudulent accounts ran about 28.8 million interactions with Claude between April 22 and June 5, 2026.

  • It claims operators linked to Alibaba's Qwen lab targeted Claude's software engineering and agentic reasoning, its most prized capabilities.

  • The June 10 letter went to Senate Banking Committee chair Tim Scott and ranking member Elizabeth Warren ahead of an AI hearing.

Anthropic Accuses Alibaba's Qwen Lab of Using 25,000 Fake Accounts to Copy Claude

Anthropic has accused Alibaba of the largest attempt yet by a Chinese company to piggyback on a leading US AI lab. The claim landed in a letter to senators and White House officials, and it has already rattled markets and sharpened the fight over how the US guards its AI work.

What Anthropic alleges

Anthropic says operators connected to Alibaba's Qwen AI lab deployed thousands of fraudulent accounts to evade its rules and access Claude, directly circumventing its policy that bars the software from being used inside China. It called the effort the largest distillation campaign yet against the company and the first time it has named a major Chinese tech conglomerate as the source.

How the alleged method works

Anthropic calls the technique adversarial distillation, where outsiders repeatedly prompt an advanced model to harvest its reasoning patterns, then use those responses to train their own software while skipping millions in research costs.

The company says the campaign could accelerate China's ability to reach the capabilities of its advanced Mythos Preview model. It also warned that models cloned this way often strip out the safety guardrails built into the originals.

Not the first time

In a February post, Anthropic flagged campaigns by DeepSeek with over 150,000 exchanges, Moonshot AI with more than 3.4 million, and MiniMax with over 13 million. Anthropic, OpenAI, and Google have since joined forces to share information about distillation attempts that break their terms of service.

Washington reacts

Senators Bill Hagerty and Andy Kim plan an amendment to defence legislation that would blacklist or sanction any Chinese firm found improperly accessing US AI output. Anthropic urged the US to clarify antitrust guidelines, reiterated support for export controls on advanced chips, and asked for penalties on firms using distillation.

The other side

Alibaba had no comment, and it was added this month to the Pentagon's list of Chinese military companies, a designation it denies and is suing to remove. Anthropic itself is under pressure, having received an export control directive suspending access to its latest Fable 5 and Mythos 5 models.

The allegations have not been independently verified, and Alibaba's US-listed shares fell more than 3% to $99.10 on the news.

What happens next depends on whether Washington treats this as a terms-of-service spat or a national security line. Anthropic clearly wants the second reading.

Read more:

https://x.com/kimmonismus/status/2069879640835961277

Quick Links:

Similar Posts