Alibaba Bans Claude Code Over Backdoor Risks, Pushes Qoder
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Alibaba banned employees from using Anthropic's Claude Code and instructed them to remove all Claude models from work computers, citing alleged backdoor security risks
- Alibaba is redirecting staff to its in-house AI coding platform Qoder, with the internal prohibition reportedly taking effect July 10
- The crackdown follows claims that Claude Code contains hidden code designed to track Chinese users — one Bluesky commentator, citing Anthropic, described it as 'admitted spyware in Claude Code designed to identify Chinese users'
- Anthropic doesn't officially serve users in China, meaning the ban implicitly confirms Alibaba had been tolerating unauthorized Claude usage internally, per @poezhao0605
- Anthropic is also pushing back from the other direction, per the Financial Times: it is moving to close loopholes that let Chinese firms such as Ant Group access its models via third-party cloud providers and overseas subsidiaries
- Kevin S. Xu noted on X that 'the list of domains and proxies to detect is LONG,' flagging Baidu, Kuaishou, ByteDance, DeepSeek and Moonshot as likely next to ban Claude Code
Why it matters: The security concerns Alibaba cites go in both directions: Alibaba is banning Claude Code at the workplace level, while Anthropic separately is shutting down overseas workarounds that let Chinese subsidiaries reach its models. For the roughly half-dozen other major Chinese AI companies named as detection targets, the precedent shifts the cost of unauthorized Western AI use from tolerated practice to a formal compliance line, and makes Alibaba's Qoder the first major in-house competitor absorbing the displaced workload.
