Cursor marketplace hosted a malicious Claude Code extension for about eleven hours on Thursday, potentially affecting up to five thousand users. Anthropic was not involved in the development of the tampered extension, which attempted to harvest sensitive user data including usernames, hostnames, operating systems, and public IP addresses. Additionally, the unauthorized extension tried to establish remote access for the attacker, although this secondary attempt likely failed on most machines because it required administrative privileges.
Upon discovering the security breach, Cursor immediately removed the malicious extension and took its entire marketplace offline to contain the threat. The company collaborated with infrastructure providers to neutralize the malware and verified that no other extensions were compromised during the incident. Following these safety checks and confirmation that there was no ongoing risk to users, the Cursor marketplace was successfully reinstated.
This incident highlights the growing security risks associated with third-party extensions in AI development environments and the rapid response required from platform operators.