Anthropic clamps down on Chinese access to Claude AI

Anthropic is tightening security to block Chinese firms from Claude AI, closing loopholes via subsidiaries, VPNs, and crypto relays. Alibaba bans Claude Code, escalating US-China AI tensions.

Anthropic is ramping up efforts to prevent Chinese companies from accessing its Claude AI models after a major distillation attack linked to Alibaba’s Qwen lab. The attack involved 25,000 fraudulent accounts and over 28.8 million interactions. To address this, Anthropic is closing loopholes exploited through overseas subsidiaries, cloud providers, VPNs, and crypto-related relay services like B.AI. Reports indicate that firms such as Ant Group and ByteDance used Singapore-based entities and VPNs to bypass restrictions, violating Anthropic’s terms of service. In response, Anthropic is tightening identity checks, monitoring usage patterns, and advocating for stronger penalties and improved information sharing among U.S. AI companies. Meanwhile, Alibaba is prohibiting employees from using Claude Code, citing alleged backdoors that could identify users connected to Chinese networks. The ban, effective July 10, 2026, requires staff to switch to Alibaba’s own coding platform. These developments underscore rising security concerns and intensifying US-China AI competition.