Compromised Fireworks Key Enables Trojanized Cursor Extension
Original source
Cursor Says Stolen Fireworks AI Key Led to Trojanized Claude Code Extension Incident
The Hacker News →A breach at Fireworks AI exposed an access key that Cursor had issued to the vendor. An attacker used this key to access a read-only storage bucket containing limited coding agent session traces from users who hadn’t enabled Privacy Mode (Cursor’s opt-out mechanism for training data). The bucket held credentials, exposing API keys for less than 0.01% of customers. The attacker then leveraged these credentials to infiltrate Cursor’s internal systems, gaining access to archived source code repositories, customer support conversations, product usage data, agent interaction summaries, and some user-supplied credentials held by the support team.
On October 29, 2026, the attacker injected trojanized versions of the Claude Code extension directly into the cache serving Cursor’s extension marketplace. The malicious builds (versions 2.1.312 and 2.1.313) attempted to install a persistent SSH backdoor. At most 5,000 users who fetched cached versions during an 11-hour exposure window potentially received the compromised extension. Cursor discovered and removed the malicious builds, took the marketplace offline, and coordinated with hosting providers to disable the attacker’s infrastructure.
The incident demonstrates how vendor compromises cascade through supply chains to downstream users. Privacy Mode users escaped impact, and no further malicious activity has been detected since November 1. The Fireworks breach also compromised other companies relying on the vendor: Uber confirmed theft of driver license information through compromised credentials, and Notion reported data theft in the same incident.
Read the full article
Continue reading at The Hacker News →This is an AI-generated summary. Read the original for the full story.