Alibaba bans staff from using Anthropic's Claude Code

Alibaba Group ordered employees to stop using Anthropic's Claude Code, an AI coding assistant, after internal reviews flagged risks tied to data security and model distillation, people familiar with the matter said. The directive, issued this week, covers all staff working on Alibaba's cloud and AI products.

The move follows heightened scrutiny over how Chinese technology companies handle user data when interacting with foreign AI tools. Alibaba's internal memo cited concerns that Claude Code could transmit proprietary code or training data to Anthropic's servers, potentially exposing the company's model architecture. The ban also addresses distillation claims – the practice of using one AI model to train another – which Alibaba views as a competitive risk.

Anthropic's Claude Code, launched in February, lets developers generate and debug code through natural-language prompts. The tool has gained traction among Chinese engineers despite restrictions on U.S. AI exports. Alibaba's prohibition mirrors similar steps by Tencent and ByteDance, which have restricted employee access to OpenAI's ChatGPT and Google's Gemini over the past year.

The ban creates a compliance burden for Alibaba's AI teams, which had integrated Claude Code into internal workflows. Engineers now must shift to domestic alternatives such as Baidu's ERNIE Bot or Alibaba's own Tongyi Qianwen model. The switch could slow development cycles in the short term, two employees said.

Alibaba's Alpha Score sits at 42 out of 100, a mixed rating that reflects the company's strong cloud revenue growth weighed against regulatory overhang and slowing e-commerce margins. The Claude Code ban adds another layer of operational friction for the AI division, which competes directly with Baidu and Tencent in the large-language-model race.

The broader implication for investors is the rising cost of compliance in China's AI sector. Each restriction on foreign tools forces domestic firms to rebuild internal tooling, diverting engineering hours from product development. Alibaba did not respond to a request for comment.