Cursor says the attack that put a malicious copy of Anthropic’s Claude Code extension in its extension marketplace began with an access key it had issued to Fireworks AI, an AI model training and inference provider. The key “was exposed by Fireworks,” the company said on November 13, and the attacker used it to copy data from a read-only storage bucket Cursor used to share data with Fireworks.
The bucket held coding agent traces, mostly from 2025, from sessions not in Cursor’s Privacy Mode, which Cursor was permitted to use for model training. A small number of the traces contained credentials that had not been redacted. Cursor said fewer than 0.01 percent of customers had API credentials exposed, and that traces were exposed for fewer than 0.05 percent of users.
Starting with those credentials, the attacker reached some of the software-as-a-service tools Cursor uses. Through them, according to Cursor, the attacker accessed some archived non-production source code repositories, limited customer support data and email conversations, and product usage metrics and contact details. The attacker also reached summaries of some coding agent conversations and some credentials users had sent to Cursor’s support team.
On October 29, using a stolen credential from the exposed data, the attacker placed tampered copies of the extension, labeled versions 2.1.312 and 2.1.313, “directly on the cache used to serve the Cursor VS Code extension marketplace,” Cursor said. Open VSX, the marketplace’s source, never carried the malware. At most 5,000 users who requested the cached versions that day may have received a malicious build, which was available for about 11 hours.
Cursor said the last attacker activity it has seen was on November 1, and that it was blocked. It said its production systems, customer code repositories, the Cursor app and its official updates, and prompts and code from Privacy Mode users were not affected. Cursor ships its app from a separate code hosting system that the attacker never accessed, and no release signing keys were exposed, the company said.
Cursor’s account does not say when or how Fireworks exposed the key. Fireworks said in an October 9 bulletin that it had identified a security incident involving unauthorized use of internal credentials, in which a third party accessed environment variables used in evaluation jobs in a limited number of customer accounts. It said it believed API keys and tokens stored as secrets had been captured.
In an update, Fireworks said it later found that account and billing information was affected, along with secrets, datasets and incomplete model weights for a limited set of accounts. It said a third-party cloud provider hosting its research and development machines had a security incident on October 23, which the intruder used the next day to reach Fireworks development machines.
Fireworks said the evidence indicates links to a financially motivated cybercriminal group that aims to sell tokens from unauthorized access to large language models to illicit routing gateways, a practice known as LLMJacking. Fireworks is among several AI developer platforms that disclosed security incidents in October, including Railway, where keys stolen in its breach ran up an estimated $740,000 in unauthorized AI usage.
Cursor has moved all inference traffic off Fireworks, deactivated the key it had issued and cut off Fireworks’ access to all the data sets Cursor had shared with it. It rotated every exposed credential it found and, as a precaution, all credentials in use across its internal systems. It also added secret scanning across its code repositories, rebuilt its VS Code extension marketplace and redeployed its production infrastructure.
Sources: Cursor, Fireworks AI (October 9), Fireworks AI (update)
–
By the Control Plane Editorial Team