Sovereign AI·Global

Anthropic Exposes Covert Chinese User Flag in Claude Code

Global AI Watch · Editorial Team··5 min read
Anthropic Exposes Covert Chinese User Flag in Claude Code
Editorial Insight

Anthropic's rollback reflects rising global scrutiny over AI privacy features, paralleling past security-data exposure cases.

Key Points

  • 1First publicly exposed surveillance feature in Claude Code.
  • 2Rollback to occur in next release after criticism.
  • 3Highlights ongoing privacy debates in AI tools.

What Changed

Anthropic implemented a covert feature within Claude Code, marking the first time such an action has been publicly exposed and rolled back. Introduced in version 2.1.91, released on April 2, 2026, this feature checked user systems for connections suggesting Chinese origins, using XOR encryption to mask detection. This has provoked user trust concerns and social media backlash, leading Anthropic to announce its removal.

Strategic Implications

The exposure and rollback signify a notable shift in the surveillance and privacy landscape of AI tools. Anthropic aimed to protect its models from unauthorized use, particularly in regions without official deployment like China. However, the backlash illustrates significant distrust among users, potentially leading to stricter scrutiny and regulatory responses to similar features in AI applications globally.

What Happens Next

Anthropic plans to remove the controversial feature quickly, with a rollback expected in the forthcoming release. As users become more vigilant, developers may face increased pressure to disclose similar checks openly. Regulatory bodies might impose newer guidelines on privacy and transparency, potentially altering how AI developers implement surveillance mechanisms.

Second-Order Effects

The incident may influence supply chain decisions for tech firms, prioritizing transparency over territorial control. Firms like DeepSeek and MiniMax, previously accused of model misuse, may exploit this controversy to argue against proprietary model protections. This could accelerate open-source alternatives or data privacy legislation impacting adjacent markets.

Free Daily Briefing

Top AI intelligence stories delivered each morning.

Subscribe Free →

Explore Trackers