Alibaba to Ban Employee Use of Claude Code Over Alleged Backdoor Concerns

Alibaba has ordered employees to stop using Anthropic's Claude Code, according to a person familiar with the order cited by Reuters, after the AI coding assistant drew scrutiny over features that can help identify China-linked users. Staff are being directed to Alibaba's in-house coding platform Qoder instead. The ban reportedly takes effect in early July.

The move deepens a spat between the two companies: Anthropic accused Alibaba in late June of illicitly extracting Claude's AI model capabilities — a distillation-attack allegation — and Alibaba's response frames Claude Code as a supply-chain and surveillance risk. Claude Code had become popular among Chinese programmers despite Anthropic's formal access restrictions on users and entities in China.

The episode is a proxy for the broader US-China AI decoupling and lands amid a Chinese pivot toward domestic and open-source models like DeepSeek, Qwen, Moonshot and Zhipu. It also arrives days after the US used export-control law to suspend Fable 5, making both governments' willingness to weaponize AI tooling starkly visible.

On Hacker News the story drew 316 points and 272 comments, stoking supply-chain-trust debate. Imbue CEO Kanjun Qiu amplified research claiming Chinese LLMs produce more vulnerable code when prompted with a US-government persona and inject PRC-aligned bias — fueling the mutual-distrust narrative. Alibaba separately backed Kuaishou's Kling AI in a $2.8B raise. Watch whether other Chinese firms follow Alibaba's ban.