Anthropic accused Alibaba of illicitly extracting capabilities from Claude in a campaign it described as its largest known attack of that kind. Key Points: Anthropic said the campaign produce
Anthropic accused Alibaba of illicitly extracting capabilities from Claude in a campaign it described as its largest known attack of that kind.
Key Points:
- Anthropic said the campaign produced more than 28.8 million exchanges with Claude.
- The company said almost 25,000 fraudulent accounts were used between Apr. 22 and Jun. 5.
- The allegation comes as U.S. officials tighten scrutiny of frontier AI access linked to China.
The U.S. AI company made the allegation in a Jun. 10 letter seen by Reuters, sent before a scheduled U.S. Senate Banking Committee hearing on artificial intelligence.
Anthropic said the operation was a “distillation” effort, a method in which a weaker model is trained on outputs from a stronger one.
The company said the campaign ran from Apr. 22 to Jun. 5, generating more than 28.8 million exchanges through almost 25,000 fraudulent accounts. Anthropic said operators affiliated with Alibaba and Alibaba Qwen, the Chinese company’s AI lab, conducted the activity.
The letter was sent to Sens. Tim Scott and Elizabeth Warren, the committee’s chair and ranking member. Alibaba did not immediately respond to Reuters’ request for comment.
Also Read:DeFi’s 2026 Selloff Deepens As Trust Breaks Faster Than TVL Recovers
Anthropic said distillation could help China move faster toward capabilities similar to its advanced Mythos Preview systems.
The company also said it supported U.S. government efforts to counter such attacks through threat-intelligence sharing with private AI companies.
The allegation fits into a wider Washington fight over AI model access, intellectual property and national security. In April, the White House accused China of stealing U.S. AI labs’ intellectual property at industrial scale.
Alibaba was added this month to the Pentagon list of Chinese military companies, a designation the company is challenging. The Commerce Department has not placed DeepSeek on a trade blacklist, despite Reuters reporting that an interagency committee viewed it as a national security risk.
Anthropic had warned in February that Chinese AI startup DeepSeek, Moonshot AI and MiniMax had tried to extract Claude capabilities. It said those earlier campaigns involved more than 150,000, 3.4 million and 13 million exchanges, respectively, before the latest Alibaba claim pushed the scale higher.
Read Next:Coinbase Leaves Binance Behind With Luxembourg MiCA Win
