Alibaba Bans All Anthropic Claude AI Tools Internally, Citing Hidden Data Surveillance Code

Alibaba has issued an internal directive requiring all employees to stop using Claude AI products from Anthropic by July 10, 2026. The ban covers mainstream models including Sonnet, Opus, and Fable, as well as agentic tools like Claude Code.

The decision follows the discovery of what developer community analysis described as a hidden detection mechanism embedded in Claude Code since April 2026. The code reportedly monitored users' device time zones and domain lists in real time, and automatically triggered a reporting action if it detected users in China time zones or accessing Chinese tech company domains — including Alibaba's.

Key Takeaways

• Hard July 10 deadline: All Alibaba employees must stop using Anthropic Claude products — including Claude Code — by July 10, 2026, per internal instructions.
• Embedded surveillance logic: Developer community reverse engineering found that since April, Claude Code included code that detected China-based users interacting with Chinese tech company domains and automatically reported that activity — a direct violation of enterprise internal R&D security expectations.
• Anthropic's rollback: Anthropic stated the mechanism was an "experimental" measure launched in March to prevent account abuse and model distillation attacks. The code was fully removed in the July 2nd version update.

The ban marks a sharp reversal for Alibaba, which had previously offered generous AI reimbursement schemes to employees — with some engineers spending hundreds of dollars per week on Claude and GPT API calls. The incident illustrates how data security concerns are now drawing hard enterprise boundaries around which AI tools organizations will permit.

🔗 Read the full article on AIBase