GoKawiilCursor says breach via Fireworks AI credential led to malicious Claude Code extension
Cursor disclosed that attackers stole an access key stored at third-party provider Fireworks AI and used it to pull data from a shared storage bucket containing historical AI coding-agent traces. Some traces had unredacted API credentials, which attackers then used to breach other Cursor systems and push malicious versions of Anthropic's Claude Code extension through Cursor's marketplace on October 29, exposing up to 5,000 developers to malware installing SSH backdoors. Cursor says fewer than 0.05% of users had traces exposed and fewer than 0.01% had credentials compromised.
GoKawiil's interpretation of the reporting above, not reported fact.
The incident shows how a single leaked credential at a third-party AI infrastructure vendor can cascade into compromise of a developer tool's marketplace and supply chain, potentially affecting thousands of downstream developers. It also highlights the risk of storing unredacted credentials in AI training data, which Cursor's own report suggests widened the blast radius beyond the initial breach point.
- Attackers used a stolen Fireworks AI key to access Cursor's shared data bucket.
- Unredacted credentials in AI coding-agent traces enabled further system compromise.
- Up to 5,000 developers were exposed to malware via a fake Claude Code extension with SSH backdoors.
Source: bleepingcomputer.com, 2026-11-14
Published there as: “Cursor confirms security breach led to malicious Claude Code extension”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.