Alibaba Bans Claude Code Over Alleged Backdoor Risks
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Alibaba banned employees from using Claude Code internally and ordered all Claude models removed from work computers over Anthropic security concerns, per sources cited by The Information
- The ban takes effect July 10, with Qoder—Alibaba's own AI coding tool—named as the replacement in an internal notice shared on X by Poe Zhao
- Alibaba flagged Claude Code as a potential backdoor, per Reuters, with X commenter Kache framing the concern as extending to any foreign AI model Alibaba doesn't control running on infrastructure it doesn't own
- Anthropic does not officially serve users in China, meaning Alibaba's ban formalized what was already tolerated-but-unauthorized internal usage that violated Anthropic's terms, per Martin Chorzempa on X
- Anthropic is concurrently moving to close loopholes that let Chinese firms like Ant access Claude via cloud providers and overseas subsidiaries, per Financial Times sources
- Bluesky and X commentary links the dispute to Anthropic's alleged admission of spyware in Claude Code designed to identify Chinese users and separate accusations that Alibaba engaged in model distillation to reverse-engineer Claude
Why it matters: Alibaba's ban formalizes an arrangement that was already a terms-of-service violation—Anthropic doesn't officially serve China—so the move admits unauthorized usage existed inside the company. The July 10 effective date plus the Qoder replacement show Alibaba preparing to cut over to its own stack, not just blocking one tool. Simultaneously, Anthropic closing cloud and subsidiary workarounds makes this a two-sided perimeter tightening, with Dare Obasanjo's claim on Bluesky that Anthropic admitted spyware targeting Chinese users turning the 'security concern' framing into a mutual accusation of espionage.
