ShieldPack SHIELD PACK
With you since 2022

Where would you like to start?

Services, updates, and ways to contact us, all in one place

Pricing and orders →

Use the bot or chat with us—whichever you prefer

ShieldPack posts

Cursor Says Stolen Fireworks AI Key Led to Trojanized Claude Code Extension Incident

About 3 min read

**Cursor’s Claude Code Extension Compromised After Fireworks AI Breach**
Cursor, the AI‑powered coding platform, announced that its official Claude Code extension on the Cursor Marketplace was hijacked after an access key it had issued to Fireworks AI was stolen. The key was exposed during a separate Fireworks AI breach, and the attacker used it to copy data from a read‑only storage bucket that Cursor shared with the company. The compromise was discovered nearly two weeks after a trojanized version of the extension had already been distributed as a legitimate update, attempting to install a persistent SSH backdoor on users’ machines.
The read‑only bucket contained traces of coding agent sessions that were not in Cursor’s “Privacy Mode.” Privacy Mode is a setting that lets users opt out of having their data used for training. Most of the sessions in the bucket dated to 2025 and were allowed to be used for model training. Among the traces were credentials that had not been scrubbed, exposing the API keys of fewer than 0.01 % of Cursor’s customers. The attacker leveraged these credentials to reach select software‑as‑a‑service tools used by Cursor and then accessed additional data.
In response, Cursor deactivated the vendor’s key, revoked and rotated all related credentials, and conducted a hunt for deeper access. The company also temporarily disabled internal and product services, which disrupted access to some Cursor features. On October 29, 2026, the attacker used a stolen credential from Cursor’s internal system to place trojanized versions of the Claude Code extension directly on the cache that serves the marketplace. Up to 5,000 users who requested specific cached versions of the extension on that day may have received the malicious build. The build was available for roughly 11 hours before Cursor removed it, took the marketplace offline, and worked with the cloud provider hosting the attacker’s server to shut it down.
Cursor Says Stolen Fireworks AI Key Led to Trojanized Claude Code Extension Incident
Cursor notified users who installed versions 2.1.312 or 2.1.313 during the exposure window and assisted them in revoking exposed credentials. The company also moved all inference traffic off Fireworks, deactivated the key issued to Fireworks, and cut off its access to all datasets shared with it. No additional malicious activity has been detected since November 1, 2026.
Importantly, there is no evidence that the incident affected users who had enabled Privacy Mode, customer code repositories, Cursor’s own production environment, the Cursor app and its official updates, other apps, or any other extension or plugin downloaded from the marketplace.
The incident unfolded against a backdrop of other breaches linked to Fireworks AI. Uber confirmed that unknown threat actors stole thousands of driver‑license records after compromising internal credentials of a service provider. On October 9, 2026, the provider notified Uber that an unauthorized third party had accessed Uber’s account with the provider, and by October 13, 2026, the investigation showed that the compromised credentials had been used to access Uber data connected to the provider. TechCrunch reported that the provider was Fireworks AI, though Uber has not disclosed how many users were impacted. Notion also confirmed that its data was stolen in the same Fireworks AI hack last week.
For ShieldPack readers, the key takeaways are that a single exposed key can

Source: The Hacker News · Cybersecurity

ShieldPack media