Alibaba has banned employees from using Anthropic’s Claude Code after developers discovered a hidden backdoor that could detect whether users were connected to China. Anthropic confirmed the backdoor was real, saying it was added in March to detect unauthorized resellers and help prevent AI model theft through “distillation.” The company said it looked at signals like time zones, proxy settings, and other indicators to identify suspicious access. Alibaba has now classified Claude Code as high-risk software internally and is telling employees to switch to its own AI coding assistant, Qoder.

Alibaba has reportedly told employees to uninstall and stop using Anthropic products, including Claude models and Claude Code, with the internal ban taking effect on July 10. The stated reason: security risk. Chinese reports say Alibaba has classified Claude Code as high-risk software after allegations that recent versions contain hidden mechanisms to identify China-based users via timezone, proxy/API routing, and subtle punctuation changes in system prompts. The timing is hard to ignore. Anthropic recently accused Alibaba/Qwen of using roughly 25,000 fake accounts to run 28.8M Claude interactions between April 22 and June 5, calling it an “industrial-scale model distillation” campaign.

Alibaba Cloud Claw Talks EP6 will be live on June 24 at 10:00–10:30 AM (UTC+8)! Let's explore Data + Agent = Your AI Workforce: Launching AI-native ApsaraDB Enterprise Agents — autonomous, secure, self-learning AI inside the database. Register now & see the future of data ops!

Alibaba Cloud at SAP Sapphire Madrid Join Douglas Wang and Liting Zhou to explore the AI-powered flywheel for global enterprises and SAP operations. 🗓 May 21 | 11:30 AM CEST 📍 Hall 9, Theater 3 🔗: