Anthropic has accused Alibaba of orchestrating the largest attempt yet to clone its Claude AI model, allegedly using 25,000 fraudulent accounts to generate over 28.8 million exchanges between April 22 and June 5. The Chinese tech giant reportedly employed obfuscation techniques and proxy networks to bypass access restrictions and mine Claude's most valuable capabilities, including agentic reasoning and software engineering.

These attacks occurred after Trump issued warnings against foreign AI distillation efforts, despite Alibaba maintaining NYSE listing and significant US business operations. Previously, Anthropic identified similar campaigns by Deep Seek, Moonshot, and Mini Max involving approximately 24,000 accounts and 16 million exchanges. OpenAI and Google have documented comparable attempts targeting their models.

Anthropic recommends Congress pass legislation enabling AI firms to share intelligence on Chinese tactics, impose stricter chip export controls, and penalize labs engaging in distillation attacks. Such measures could limit Chinese access to US models and advanced computing resources. Alibaba's Qwen models have reportedly been downloaded over 700 million times.

The dispute escalates as Alibaba sues the Trump administration over blacklisting claims, denying military ties while China races to develop competitive AI capabilities. Alibaba's stock fell 3 percent following the public accusations, highlighting investor concern about US-China AI tensions.