Updated
Updated · Microsoft · Oct 30
Microsoft Finds SSH Backdoor in Claude Code 2.1.313 on Cursor, Exposing Up to 5,000 Users
Updated
Updated · Microsoft · Oct 30

Microsoft Finds SSH Backdoor in Claude Code 2.1.313 on Cursor, Exposing Up to 5,000 Users

3 articles · Updated · Microsoft · Oct 30

Summary

  • Microsoft said the trojanized Claude Code extensions on Cursor—versions 2.1.312 and 2.1.313—loaded injected JavaScript that tried to add an attacker SSH key, enable local SSH, open a reverse tunnel and persist across reboots.
  • The code also gathered host details including hostname, username, OS, Git email and up to 40 environment-variable names matching terms like KEY, TOKEN, SECRET, AWS or GCP, then attempted to beacon them to 134.209.82.225:8099.
  • Cursor said the malicious packages were live from about 2:27 a.m. to 1:35 p.m. PDT on Oct. 29, while Microsoft said Visual Studio Marketplace and npm were unaffected and Anthropic's own systems were not compromised.
  • Microsoft warned that failed tunnel setup did not mean a clean device, because the loader could still plant SSH keys and persistence artifacts such as com.apple.sysmond.plist, .smt scripts, SysMon tasks or Startup-folder files.
  • Organizations are being told to hunt for installs from Cursor, preserve evidence, remove unauthorized keys and startup items, and rotate credentials or rebuild devices if telemetry shows exposure.

Insights

How did a malicious Claude Code update slip through Cursor’s official marketplace and potentially backdoor 5,000 developer machines in just hours?
If you installed Anthropic’s Claude extension, could uninstalling it still leave an SSH backdoor, stolen secrets, or hidden persistence behind?
What does this attack reveal about the hidden risks of auto-updating trusted AI coding extensions with IDE-level access?