
Cursor, the AI code editor, has published its incident report on the October breach that put a malicious copy of Anthropic’s Claude Code extension in its extension marketplace. The company says it began at Fireworks AI, a vendor Cursor used for model training and inference: an access key Cursor had issued to Fireworks was exposed there, and the attacker used it to copy data from a read-only storage bucket the two companies shared. The bucket held historical coding-agent traces, mostly from 2025, from sessions not in Privacy Mode. A small number of those traces contained credentials that had not been redacted.
Cursor says the attacker used those credentials to reach select SaaS tools and take archived non-production source code repositories (some with business contact information), limited customer support data and email conversations, product usage metrics and contact details, summaries of some coding-agent conversations, and some credentials users had sent to Cursor’s support team. A credential from that data was then used on October 29 to place the malicious extension copies directly on the cache that serves the Cursor marketplace. Cursor says the extension was up for about 11 hours, that at most 5,000 users who requested the cached versions may have received it, and that Anthropic did not publish it and the Open VSX marketplace never had the malware. Its figures: coding-agent traces were exposed for fewer than 0.05% of users, and API credentials for fewer than 0.01% of customers. The company says the last attacker activity it saw was on November 1 and was blocked, and that it believes the incident is contained.
The report names two tampered builds, 2.1.312 and 2.1.313. Our October 30 post on the backdoor cited only 2.1.313; Cursor’s notice, whose title now lists both, is the reference. Cursor says production infrastructure, the Cursor app and its official updates, release signing keys, customer code repositories, and the prompts and code of Privacy Mode users were not affected. It says it has moved all inference off Fireworks, deactivated the key and cut Fireworks’ access to shared data sets, rotated every exposed credential it found (and, as a precaution, all credentials in use internally), rebuilt the marketplace and redeployed production infrastructure. BleepingComputer, which reported the update, ties the incident to what Wiz calls “Black October”, a run of attacks on AI developer platforms. Our November 11 post covers Fireworks’ own disclosure of two breaches; Cursor’s report does not say which Fireworks incident exposed its key.
How to check if you’re affected
Affected versions of the extension are Anthropic.claude-code 2.1.312 and 2.1.313, if you installed or updated to either from the Cursor marketplace during the October 29 exposure window. Cursor asks those users to follow the steps in its security notice, and says uninstalling the extension alone may not be enough. The checks and clean-up scripts are in that notice; our October 30 post summarizes the SSH-key check from it.
For the data exposure, Cursor says it has identified affected users and notified each of them directly, and that if you did not receive a notification, your data was not in the exposed traces. It says it emailed team admins a list of specific exposed keys to rotate, and asked the services that issued them to revoke them. If you ever sent a credential to Cursor’s support team, rotating it is a cautious step on our part; the report says only that some such credentials were taken.
Sources
- Cursor: An update on our recent security incident
- Cursor: Security notice, malicious version of Anthropic.claude-code (2.1.312 and 2.1.313) identified in the Cursor VS Code Extension Marketplace
- BleepingComputer: Cursor confirms security breach led to malicious Claude Code extension
- Wiz: “Black October” for AI Developer Platforms (linked from BleepingComputer; the Wiz page was not read in full)
