Cursor links malicious Claude Code extension to Fireworks AI breach
AI summary
Cursor said an unauthorized, modified Anthropic Claude Code extension was listed in its marketplace on October 29 after credentials stolen through a Fireworks AI breach were used to access Cursor systems. The malicious versions remained available for about 11 hours and may have reached as many as 5,000 users; Cursor removed them, rotated credentials, notified affected customers, and said it found no attacker activity after November 1.
In-depth insight
The investigation shows how a breach at an AI infrastructure provider enabled a software-supply-chain attack against an AI coding tool, exposing the security of extension marketplaces, stored credentials, and non-privacy-mode coding sessions.