Skip to main content

Alibaba Bans Claude AI Tools After Hidden Detection Found

Alibaba has issued an internal directive banning all employees from using Anthropic's Claude series of AI tools, including popular models like Sonnet, Opus, and Fable, as well as the intelligent agent tool Claude Code. The deadline for compliance is July 10.

This marks a sharp reversal from earlier this year, when Alibaba actively encouraged AI adoption. The company had provided free quotas for its own models and offered generous reimbursement for external ones. Employees quickly embraced tools like Claude and GPT, with some programmers spending hundreds of dollars per week on API calls. But as concerns over data security and AI computing power grew, the company pulled the plug.

Image

The ban stems from a hidden detection mechanism discovered in Claude Code. According to reports from the developer community, since April, the tool has included code that monitors users' device time zones and the domains they access. If it detected a user in the China time zone or interactions with domains belonging to Chinese tech companies—including Alibaba—the system would automatically report the activity. This practice of embedding sensitive detection logic directly into the system prompt crossed a red line for enterprise security.

Anthropic responded by stating that the mechanism was an "experimental" measure launched in March to prevent account abuse and model distillation attacks. The company said the relevant code has been completely rolled back and deleted in the latest version released on July 2.

Key Points

  • Alibaba bans all Claude series AI tools internally, effective July 10.
  • The ban follows discovery of a hidden detection mechanism in Claude Code.
  • The mechanism monitored time zones and domain access, flagging Chinese tech interactions.
  • Anthropic says the feature was experimental and has been removed.