
AI-powered code editor Cursor has officially confirmed a security breach that allowed cybercriminals to publish a malicious Claude Code extension directly on its official marketplace. According to the company’s incident report, the attackers gained unauthorized publishing access after stealing valid developer credentials during an earlier third-party cloud breach at AI infrastructure provider Fireworks AI. Software engineers who installed the trojanized extension risked exposing local source code, environment variables, and API secrets. Cursor has since revoked the compromised tokens, purged the rogue extension, and urged developers to rotate any sensitive credentials immediately.