Anthropic has accused Alibaba of carrying out the largest known attempt to clone its Claude AI model, alleging the use of nearly 25,000 fraudulent accounts to extract millions of interactions from the chatbot. The company claims this activity was aimed at replicating Claude's advanced capabilities, including agentic reasoning and long-horizon task execution, without incurring the cost of developing a new model from scratch.

According to a confidential letter sent by Anthropic to US Senators Tim Scott and Elizabeth Warren, operators linked to Alibaba and its AI division, Qwen, allegedly generated over 28.8 million exchanges with Claude between April 22 and June 5. This activity, the company claims, violated its terms of service and was designed to illicitly extract Claude’s most advanced features.

Anthropic describes the incident as the largest distillation attack it has ever detected. Distillation involves querying an advanced AI model to reproduce its capabilities in another system without building a new model from the ground up. The company alleges that the campaign relied on proxy networks and obfuscation techniques to avoid detection.

The allegations highlight growing concerns over AI model theft and intellectual property violations in the global tech sector. They raise questions about the cost of developing advanced models, the risks of vendor lock-in, and the need for stronger governance frameworks to protect proprietary AI systems. Market reactions could influence regulatory responses and industry standards for AI security.

The situation underscores the intensifying competition between the US and China in the field of artificial intelligence. As the US government considers potential actions against Alibaba, the case may set a precedent for how intellectual property violations are addressed in the rapidly evolving AI landscape.