Top stories
Alibaba to ban Claude Code at work over alleged backdoor risk after Anthropic confirms a proxy check
- Category: Agentic coding
- Status: developing
- Sources: Reuters, HN discussion
- Summary: Reuters reported on 2026-07-03, citing a source, that Alibaba will bar employees from using Claude Code in workplace environments starting 2026-07-10, after Chinese financial outlet Yicai reported Alibaba identified an embedded backdoor risk in the tool. The allegation originated in a 2026-06-30 reverse-engineering writeup claiming that Claude Code since v2.1.91 (2026-04-02) silently inspected users' proxy configuration and system time zone. A member of Anthropic's Claude Code team said on social media that the mechanism existed to detect account resale and model distillation rather than to spy on users, and that it would be removed in the next update. No third-party security firm has independently confirmed a backdoor.
- Comments: HN commenters tied the ban to the request-marking and proxy-check reports of the prior days and argued the incident is a reason to prefer open-source coding agents and local models; others stressed that the reverse-engineering claims remain independently unverified.
- Why it matters: A major cloud vendor banning a widely used coding agent, together with Anthropic acknowledging an undisclosed environment check, turns the running Claude Code telemetry question into a concrete tool-selection and data-governance decision for engineering teams.
- Follow-up: Track the Claude Code update that removes the proxy and time-zone check, any Anthropic statement or documentation change, and whether other firms restrict the tool.