Cursor Says Stolen Fireworks AI Key Led to Trojanized Claude Code Extension Incident

The Hacker News
Read full story· 4 min
Cursor disclosed a data breach caused by a stolen Fireworks AI access key, leading to exposure of limited user data and internal source code. The attacker used stolen credentials to distribute trojanized versions of Anthropic's Claude Code extension via Cursor's marketplace, affecting up to 5,000 users.

What happened

Cursor reported a security breach after an access key it had issued to Fireworks AI was stolen. The attacker used this key to access a read-only storage bucket containing limited user data and internal source code. On October 29, 2026, the attacker uploaded trojanized versions of Anthropic's Claude Code extension to Cursor's extension marketplace, which were available for about 11 hours. Cursor has since removed the malicious extension, notified affected users, and taken steps to secure its systems.

Key details

  • Up to 5,000 users may have downloaded the trojanized Claude Code extension on October 29, 2026.
  • The malicious extension attempted to install a persistent SSH backdoor on affected systems.
  • The breach exposed some API credentials for less than 0.01% of Cursor customers.
  • Only sessions not in Privacy Mode were affected, most dating back to 2025.
  • Cursor disabled the compromised vendor key, rotated credentials, and took its extension marketplace offline during the incident.

Why it matters

Cursor users who downloaded specific versions of the Claude Code extension on October 29, 2026, face the risk of having their systems compromised by a backdoor. A small number of customers had their API credentials exposed, requiring them to revoke and replace those credentials. Cursor had to temporarily disrupt some of its services to contain the breach, affecting user access and extension availability.


More in Cybersecurity