Security Incident Affects Claude Code Extension via Cursor Marketplace
A security breach involving a trojanized version of the Claude Code VS Code extension has been identified, targeting users who download the software through the Cursor IDE extension marketplace. The incident involves a malicious update, specifically version 2.1.313, which was distributed under the extension’s legitimate publisher identity.
Nature of the Compromise
Cybersecurity researchers have disclosed that the compromised extension was designed to install a persistent SSH reverse backdoor on infected hosts. This backdoor could potentially allow unauthorized remote access to the user’s system. While some unverified reports on social media platforms like Reddit have characterized the event as a direct compromise of Anthropic, the developer of the Cursor IDE has provided a different clarification.
In a statement released via X (formerly Twitter), Cursor confirmed a security incident affecting a modified Claude Code extension accessed through its specific marketplace. The company stated that the malicious version was not developed by Anthropic and estimated that at most 5,000 users were impacted. Cursor further clarified that the breach does not affect other extensions downloaded from the cursor.com marketplace, nor does it impact other products such as Grok Bot or Grok.
Scope and Mitigation
The breach appears to be localized to the Cursor marketplace rather than the broader Open VSX registry or the official Microsoft VS Code marketplace. Consequently, users who do not utilize the Cursor IDE may not be at risk.
Following the discovery of the unauthorized update, Cursor temporarily shut down its marketplace to remove the compromised package. The exact method used by attackers to bypass authentication and introduce the unauthorized update remains unclear.
Industry Reaction
The incident has prompted calls for greater transparency from developers. Some users on community forums have expressed concern regarding the potential for remote access trojans and have called for a formal explanation from Anthropic, noting the high-level access such a backdoor provides to developer laptops.
At present, the incident remains a targeted breach affecting a specific subset of the developer community using the Cursor ecosystem. No direct link has been established between this breach and the core security infrastructure of Anthropic’s LLM services.