AI-native code editor Cursor has published a detailed account of the security breach that led to a malicious copy of Anthropic’s Claude Code extension appearing in its VS Code extension marketplace late last month, tracing the attack back to an access key that was exposed by its third-party inference vendor, Fireworks.
The company said the vast majority of its customers were not affected and that it has contacted each impacted user directly. The last unauthorized activity Cursor observed was on November 1, and it was blocked.
How the attack unfolded
The breach began when an attacker obtained an access key that Cursor had issued to Fireworks, the vendor it used for model training and inference. Fireworks disclosed the key exposure in its own security bulletin. The attacker used the key to copy data from a read-only storage bucket Cursor had shared with Fireworks, which contained limited coding agent traces, mostly from 2025, collected from sessions where Privacy Mode was not enabled.
A small number of those traces contained credentials that had not been properly redacted. Cursor said fewer than 0.01% of its customers had API credentials exposed this way. The attacker then used those stolen credentials to reach several SaaS tools used internally by Cursor, gaining access to archived non-production source code repositories (some containing business contact information), limited customer support data and email conversations, product usage metrics and contact details, summaries of some coding agent conversations, and credentials that users had sent to Cursor’s support team.
Using a credential stolen from that internal data, the attacker placed malicious, unauthorized copies of the Claude Code extension directly on the cache used to serve the Cursor VS Code extension marketplace. Cursor stressed that Anthropic did not publish the malicious copy and that this was not a vulnerability in Anthropic’s product. The upstream marketplace source, Open VSX, was never compromised.
Cursor estimates that up to 5,000 users who requested specific cached versions of the extension on October 29 potentially received the tampered build, which was available for around 11 hours before it was removed. Cursor took its extension marketplace offline entirely and worked with the cloud provider hosting the attacker’s server to shut it down.
What was not affected
The company emphasized several areas the breach did not reach. Prompts and code from users with Privacy Mode enabled were not exposed, and coding agent traces were exposed for fewer than 0.05% of users overall. Production systems, the Cursor app itself, and official software updates were unaffected, as those are shipped from a separate code hosting system the attacker never accessed. No release signing keys were exposed.
Active inference requests, customer code repositories, plugins downloaded from the Cursor Marketplace, and all other extensions in the VS Code marketplace were also unaffected. Cursor said it validated every other extension for authenticity and rebuilt the marketplace store.
Cursor said it moved all inference traffic off Fireworks, deactivated the compromised key, and cut off the vendor’s access to all shared data sets. The company rotated every exposed credential it identified, prioritizing the highest-risk ones first, and as a precaution rotated all credentials across its internal systems even where there was no evidence of exposure.
Additional steps included expanding network access restrictions across service providers so that compromised credentials cannot be used outside their intended location, adding secret scanning across all code repositories and other internal data sources, redeploying all production infrastructure as a precaution, and engaging external forensics experts.
Cursor began notifying affected users on October 30 through a security notice that included indicators of compromise and remediation scripts. For customers whose credentials were exposed, team admins received lists of the specific keys to rotate, and Cursor asked the third-party services that issued those credentials to revoke them. The company warned that users who installed the tampered versions of Anthropic.claude-code (versions 2.1.312 or 2.1.313) during the exposure window should follow the steps in its security notice, as uninstalling the extension alone may not be sufficient.
Cursor said it believes the incident is now contained and continues to monitor its systems for anomalous activity. Questions can be directed to [email protected].

Leave a Reply