Anthropic's Official Cursor Extension Backdoored in Supply-Chain Attack Affecting 5,000 Developers

A compromised version of Anthropic's Claude extension on the Cursor marketplace distributed malicious code targeting cloud credentials.

An official extension for Anthropic's Claude model distributed via the Cursor AI development environment was compromised in a supply-chain attack, exposing up to 5,000 developers' machines to potential credential theft. The malicious version 2.1.313 was pushed directly through official distribution channels.

Security researchers and industry experts flagged the unauthorized backdoor, which actively hunted for cloud credentials before Anthropic deployed a remediation patch with version 2.1.314 hours later. The incident highlights critical vulnerability vectors within AI-assisted development tools and third-party marketplace supply chains.