Cursor Confirms Breach Led to Malicious Claude Code Extensions

· Original article ↗

Summary

Cursor says attackers used a stolen Fireworks AI key to access internal data and credentials, then distributed malicious Claude Code extensions for about 11 hours. Up to 5,000 developers may have received builds that attempted to install SSH backdoors.

Key points

  • Cursor says attackers obtained a Fireworks AI access key and used it to copy historical coding-agent traces; some contained unredacted API credentials.
  • Using exposed credentials, attackers accessed additional Cursor systems and stole data from several SaaS tools, including limited support information, usage metrics, and customer-shared credentials.
  • On October 29, attackers used a stolen credential to publish malicious Claude Code extension versions 2.1.312 and 2.1.313 through Cursor's marketplace; the builds were available for about 11 hours.
  • The malicious extensions attempted to collect system information, install an attacker-controlled SSH key, enable SSH access, and establish persistence.
  • Cursor estimates up to 5,000 users who requested affected versions may have received them. Researchers found coding errors that likely prevented the backdoor from working on many Windows and macOS systems where SSH was not already enabled.
  • Cursor says it revoked and rotated credentials, moved inference traffic away from Fireworks, rebuilt its extension marketplace, and redeployed production infrastructure. Customers who installed or updated to an affected version are advised to follow Cursor's remediation guidance.

Article Details

Victim Organization
Cursor
Incident Type
Security breach involving credential theft, unauthorized access and data theft, followed by distribution of malicious Claude Code extensions through Cursor's marketplace.
Incident Date
2026-10-29
Data Types Exposed
  • Historical AI coding-agent traces, some containing unredacted API credentials
  • Archived, non-production Cursor source code repositories, some containing business contact information
  • Customer support information and email conversations
  • Product usage metrics and customer contact details
  • Summaries of some AI coding-agent conversations
  • Credentials customers had previously shared with Cursor's support team
Affected Records
Cursor estimated that up to 5,000 users who requested affected extension versions during the exposure window may have received malicious builds. Cursor said fewer than 0.05% of users had coding-agent traces exposed and fewer than 0.01% of customers had API credentials compromised.
Operational Impact
Attackers accessed internal systems and stole data from several Cursor SaaS tools, then used a stolen credential to publish malicious copies of Anthropic's Claude Code extension. The extensions were available for approximately 11 hours. Cursor temporarily turned off some internal and product services, removed the extensions, rotated credentials, rebuilt its marketplace, and redeployed production infrastructure. The malicious code attempted to install persistent SSH backdoors; researchers found coding errors that likely prevented the backdoor from working on many Windows and macOS systems where SSH was not already enabled.
Claim Status
confirmed

MITRE ATT&CK

Vendors

Products

Related Articles