The recent compromise of Anthropic’s official Claude Code extension on the Cursor code editor’s extension marketplace has been officially linked to a security breach at third-party AI provider Fireworks AI, which resulted in the theft of an exposed access key. The incident, which sent ripples through the software development and artificial intelligence communities late last month, has exposed the cascading risks associated with modern third-party software dependencies and shared cloud infrastructure. According to updates provided by the AI-powered coding platform Cursor, the security breach began when unauthorized actors gained access to credentials that had been improperly handled or inadvertently exposed following the earlier incident at Fireworks AI. Read Also: Cisco Warns of Active Attacks Exploiting Critical Zero-Day Flaw in Secure Email Gateways Russian Enterprises Target of Three Distinct Threat Clusters Deploying Custom Backdoors and Ransomware "An attacker obtained an access key we had issued to Fireworks," Cursor explained in an update regarding the security incident. "The key was exposed by Fireworks, and the attacker used it to copy data from a read-only storage bucket we used to share data with them." This disclosure follows by nearly two weeks the initial discovery of a trojanized version of Anthropic’s official Claude Code extension. The malicious build had been surreptitiously distributed through the Cursor Marketplace disguised as a legitimate software update, with the underlying intent of installing a persistent Secure Shell (SSH) backdoor onto the systems of unsuspecting developers and organizations that downloaded the package. Anatomy of the Breach: From Read-Only Storage to Internal Systems According to details shared by the artificial intelligence coding platform, the read-only storage bucket targeted by the attacker contained limited coding agent traces derived from development sessions that were not operating under Privacy Mode. Privacy Mode is a specific user-controlled setting within Cursor that allows developers to explicitly opt out of having their session data harvested, processed, or utilized for broader model training purposes. The affected sessions, the vast majority of which dated back to 2025, were legally permitted to be utilized for model training under standard usage agreements. However, these raw traces inadvertently contained unredacted credentials. The oversight exposed the API credentials of a small fraction of Cursor’s total user base—specifically, less than 0.01% of customers. Rather than stopping at the exposed storage bucket, the malicious actor utilized these harvested credentials as a strategic foothold. Cursor stated that the attacker leveraged the stolen information to pivot toward select software-as-a-service (SaaS) tools utilized internally by the company, thereby gaining broader access to internal data systems. Recognizing the severity of the unauthorized access, Cursor initiated emergency containment protocols. The company deactivated the compromised vendor key, systematically revoked and rotated corporate credentials, and initiated an extensive forensic investigation to hunt for deeper indicators of compromise. To prevent further lateral movement across its corporate network, Cursor temporarily disabled several internal and product-related services, causing temporary disruptions to specific features relied upon by its user base. The Marketplace Supply Chain Attack The escalation of the security incident culminated on October 29, 2026. Armed with a stolen credential obtained from the exposed internal system data, the attacker managed to place trojanized versions of Anthropic’s Claude Code extension directly into the software cache utilized to serve updates for Cursor’s extension marketplace. The malicious update was engineered to compromise developer workstations by establishing a persistent backdoor, posing a severe threat to enterprise software integrity and source code security. "At most 5,000 users who requested specific cached versions of the extension on October 29 potentially received the malicious build," Cursor noted in its incident assessment. "It was available for approximately 11 hours before we removed it. We took it down, took the extension marketplace offline, and worked with the cloud service provider hosting the attacker’s server to take it down." In the wake of the discovery, Cursor moved quickly to notify all affected users who had inadvertently installed versions 2.1.312 or 2.1.313 of the extension during the narrow exposure window. The company provided direct assistance to those individuals to help revoke and replace any exposed credentials. Furthermore, Cursor severed its technical ties with Fireworks AI by moving all model inference traffic completely off the platform, permanently deactivating the vendor key, and cutting off all data-sharing pipelines. Cursor emphasized that comprehensive monitoring and security audits have revealed no additional malicious activity since November 1, 2026. The platform also reassured its broader user base that the incident did not impact developers who had enabled Privacy Mode, nor did it affect customer code repositories, Cursor’s core production environment, the main Cursor desktop application or its official self-updates, any other applications developed by the company, or third-party extensions and plugins downloaded from the marketplace that were unrelated to the compromised Claude Code build. The Widening Impact of the Fireworks AI Incident The revelation regarding Cursor’s infrastructure highlights a much broader wave of cyber security events stemming from the Fireworks AI breach. The incident has impacted multiple prominent technology companies that rely on the provider for artificial intelligence model inference and hosting services. In a concurrent development, ride-sharing giant Uber confirmed that unknown threat actors managed to steal sensitive personal information belonging to thousands of drivers, including driver’s license records, after obtaining internal service provider credentials. "On October 9, 2026, a service provider notified Uber that an unauthorized third-party had compromised the service provider’s internal credentials and used them to access Uber’s account with the service provider," the company stated in an official data breach disclosure filing submitted to the Office of the Attorney General in California. "On October 13, 2026, our investigation determined that the unauthorized party had used the compromised credentials to access Uber data connected to the service provider." Although Uber initially withheld the specific identity of the third-party vendor in its regulatory filings, investigative reports by technology publication TechCrunch identified the organization in question as Fireworks AI. Uber has not publicly disclosed the exact number of drivers whose personal data was accessed during the breach. The incident at Fireworks AI has also affected other major pillars of the software and productivity ecosystem. Weeks prior to the disclosures by Cursor and Uber, online collaboration and productivity startup Notion confirmed that proprietary corporate data was similarly stolen as a direct result of the security lapses at Fireworks AI. As investigations into the interconnected breaches continue across the technology sector, the events underscore the vulnerability of modern software supply chains, where a single compromised vendor or an improperly secured cloud storage bucket can trigger a domino effect impacting millions of users and top-tier global enterprises. Post navigation Italy Fines Meta €805 Million Over Android Photo-Sharing Feature and Default Data Access