Alibaba Bans Anthropic’s Claude Code for Employees

Overview

• Alibaba has told staff to stop using Claude Code for work and to adopt its in‑house coding tool Qoder, with the ban set to take effect on July 10.
• Anthropic has acknowledged the March experiment that inserted a hidden session marker in Claude Code, said it will remove the marker, and is deploying stronger anti‑abuse mitigations as it restores access.
• Security researchers reported the marker detected proxies and China‑linked identifiers and transmitted the results covertly by altering date and punctuation formats in telemetry, a technique that raised privacy and trust concerns.
• The episode followed U.S. agency scrutiny and temporary export limits on Anthropic models that were later lifted by the Department of Commerce, and it has accelerated corporate efforts to limit foreign cloud tools.
• The move is likely to speed enterprise shifts to auditable domestic models and downloadable open weights, affect developer workflows inside firms such as Alibaba, and sharpen the commercial and national‑security contest between U.S. and Chinese AI stacks.