By Hamza Faisal
Cursor has closed its two-week investigation into the security incident that pushed a tampered copy of Anthropic's Claude Code extension to as many as 5,000 users, and traced the breach back to a single access key its artificial intelligence vendor Fireworks had exposed.
The code editor maker published its findings on Friday, saying an attacker obtained a key Cursor had issued to Fireworks, used it to copy data from a shared storage bucket, and used unredacted credentials in that data to widen the intrusion. The last attacker activity Cursor detected was on November 1, and it was blocked.
A leaked vendor key opened the door
According to Cursor, the stolen key unlocked a read-only bucket the company used to share data with Fireworks, its training and inference vendor. The bucket held limited coding agent traces, mostly from 2025 sessions that were not in Privacy Mode and that Cursor was permitted to use for model training.
A small number of those traces contained credentials that had not been redacted. Cursor said fewer than 0.01 percent of customers had API credentials exposed, and traces were exposed for fewer than 0.05 percent of users. Using those credentials, the attacker reached select software tools Cursor uses internally, and from there accessed archived non-production source code repositories, limited customer support data and email conversations, usage metrics, contact details, and summaries of some coding agent conversations.
Cursor said it deactivated the vendor key as soon as it spotted the unauthorised activity and began revoking and rotating credentials. Out of caution, it temporarily disabled internal and product services well beyond what the evidence required, briefly disrupting some features, before restoring each one once it was safe.
The poisoned extension reached 5,000 users
On October 29, the attacker used a stolen credential to place malicious copies of Anthropic's Claude Code extension, versions 2.1.312 and 2.1.313, directly on the cache serving the Cursor VS Code extension marketplace. The tampered builds, which carried an SSH backdoor, were available for about 11 hours before Cursor removed them. At most 5,000 users who requested the specific cached versions on October 29 may have received the malicious build.
Cursor stressed that Anthropic did not publish the copy and that the incident was not a vulnerability in Anthropic's product. The upstream marketplace source, Open VSX, never hosted the malware. Cursor took the marketplace offline, rebuilt it, and worked with the cloud provider hosting the attacker's server to take it down.
The company published a security notice on October 30 with indicators of compromise and remediation steps, and began notifying users and team administrators. It warned that simply uninstalling the tampered extension may not be enough for anyone who installed it during the exposure window.
What was spared, and what changes now
Cursor said the breach did not touch prompts and code from users in Privacy Mode, live traffic, production systems, customer code repositories, official app updates, or release signing keys. Active inference requests were not exposed, and no other extension in the marketplace was affected.
As part of its response, Cursor said it has moved all inference traffic off Fireworks and cut off the vendor's access to every shared dataset. It rotated every exposed credential and, as a precaution, all credentials in use across its internal systems. It redeployed all production infrastructure, expanded secret scanning across its repositories, and brought in external forensics experts.
“We apologize to those impacted by this incident — privacy and security is treated as a top priority, and we can and must do better,” the company said in its update.



