Reported by 2 outlets

Attackers breach Cursor via Fireworks AI to distribute backdoor-laden Claude Code add-on

Cybersecurity illustration · BleepingComputer

The story so far

AI summary of the reporting below · Updated · How we work

Cursor suffered a security breach after hackers obtained an access key stored on third-party provider Fireworks AI, allowing attackers to distribute malicious copies of Anthropic's Claude Code extension through Cursor's marketplace on October 29. The compromised extension versions, 2.1.312 and 2.1.313, were available for about 11 hours and exposed up to 5,000 developers to malware designed to install persistent SSH backdoors. [1][2]

Some traces contained unredacted credentials that granted access to internal SaaS tools and marketplace caches. Cursor noted that fewer than 0.05% of users had traces exposed and fewer than 0.01% had API credentials compromised. Researchers noted coding errors likely stopped the backdoor on many Windows and macOS machines lacking enabled SSH, and Cursor has since shifted inference traffic away from Fireworks and rebuilt its marketplace cache. [2]

Cybersecurity firm Wiz identified the intrusion as part of a broader wave of developer-targeted platform attacks dubbed Black October. The campaign has hit multiple infrastructure and AI firms, including Railway, Northflank, and Lightning AI, often seeking authentication secrets to run unauthorized model inference. [2]

Read the original reporting

2 outlets

Related coverage

In this story