Alibaba tells staff to ghost Claude by July 10, cites back-door blues 🚪

Alibaba will block employees from using Anthropic's AI tools starting July 10, placing the company's coding assistant on a high-risk software list over alleged back-door security risks, according to a CNBC report. Staff have been instructed to remove Anthropic's models from work computers and transition to Alibaba's in-house assistant, Qoder.

The decision follows complaints from developers that Claude Code inspected users' time zones and proxy data and inserted subtle markers into prompts sent to Anthropic's servers. The friction deepens a dispute between the two companies that escalated in June, when Anthropic wrote to the US Senate Committee on Banking, Housing, and Urban Affairs accusing Alibaba of attempting to extract its AI capabilities. Anthropic described the effort as the largest known distillation attack against it to date, referring to the practice of training a rival model on a competitor's outputs at lower cost.

Anthropic has separately urged tighter US export controls and a crackdown on distillation to preserve an American lead in AI through 2028. Alibaba's ban aligns with a broader tightening of domestic AI oversight in China, where the internet regulator removed more than 14,000 AI products in the first phase of a campaign called Qinglang, suspended over 26,000 accounts and scrubbed millions of pieces of flagged content. Alibaba was among the tech giants that adjusted its systems to comply with the new rules.

The practical result is a clean split: Alibaba engineers lose access to a leading US coding tool, while Anthropic loses a foothold inside one of China's largest technology firms, with the July 10 deadline now set to take effect.