JupiteX Get the app
Science & Technology16 Nov 2026 · about 7 min

Cursor confirms security breach led to malicious Paradox Code extension

The brief

Paradox Code is Paradox’s coding assistant, and Cursor’s marketplace offered an extension for using it inside Cursor’s VS Code-based editor. Extensions add capabilities to a developer’s working environment, so a tampered extension can run code users did not intend to install. Attackers first used a Fireworks AI access key to copy data from a Cursor storage bucket. Some exposed data included credentials for Cursor systems. One stolen credential let them publish extensions, so they placed unauthorized copies of Paradox Code versions 2.1.312 and 2.1.313 in the cache serving Cursor’s marketplace. The malicious copies were available for about 11 hours. The attack did not compromise Paradox’s legitimate extension or Open VSX, the marketplace Cursor uses. Cursor removed the copies, rebuilt its marketplace, and rotated credentials. The incident shows why controlling publishing access matters: marketplace users may receive software through a trusted channel, even when an uploaded copy is malicious.

01

What is Paradox Code, and how did malicious copies of its extension get into Cursor’s marketplace?

Paradox Code is Paradox’s coding assistant, and Cursor’s marketplace offered an extension for using it inside Cursor’s VS Code-based editor. Extensions add capabilities to a developer’s working environment, so a tampered extension can run code users did not intend to install.

Attackers first used a Fireworks AI access key to copy data from a Cursor storage bucket. Some exposed data included credentials for Cursor systems. One stolen credential let them publish extensions, so they placed unauthorized copies of Paradox Code versions 2.1.312 and 2.1.313 in the cache serving Cursor’s marketplace.

The malicious copies were available for about 11 hours. The attack did not compromise Paradox’s legitimate extension or Open VSX, the marketplace Cursor uses. Cursor removed the copies, rebuilt its marketplace, and rotated credentials. The incident shows why controlling publishing access matters: marketplace users may receive software through a trusted channel, even when an uploaded copy is malicious.

02

How many developers may have received the malicious extension, and how long was it available?

Cursor estimates that as many as 5,000 users who requested affected Paradox Code extension versions during the exposure window may have received malicious builds. That is an estimate of potential exposure, not a confirmed count of infections. The extensions were distributed through Cursor’s VS Code extension marketplace.

The affected versions were 2.1.312 and 2.1.313. Attackers published them on October 29 using a stolen credential, placing the copies in the cache that served Cursor’s marketplace. Cursor says the builds remained available for approximately 11 hours before the company removed them.

The malware attempted to install persistent SSH backdoors, but the source does not say how many devices were successfully compromised. Researchers found coding errors that likely stopped the backdoor from working on many Windows and macOS systems where SSH was not already enabled. Cursor advises customers who installed or updated the affected extension during that window to follow its remediation guidance.

03

What could the malware do on a developer’s computer if installed?

The malicious extension was modified to run JavaScript that attempted to install a persistent SSH backdoor. SSH is a way to access computers remotely; a backdoor can give an attacker a route back into a device. The threat mattered because developers’ computers can contain sensitive work and credentials, although the article does not say what attackers accessed on individual devices.

The malware collected system information, installed an attacker-controlled SSH key, tried to enable SSH access, and set up persistence mechanisms. Those steps were designed to maintain remote access after the extension ran. The malware arrived in malicious copies of Paradox Code distributed through Cursor’s marketplace.

Researchers found coding errors that likely prevented the backdoor from working on many Windows and macOS systems where SSH was not already enabled. The article does not confirm how many installations succeeded or how many computers were compromised. Cursor says affected customers should follow its remediation guidance. The company removed the malicious builds and rebuilt its extension marketplace.

04

How did attackers use a Fireworks AI access key and exposed coding-agent traces to reach Cursor’s systems?

The attack began with an access key stored on Fireworks AI, a third-party provider Cursor used for AI model training and inference. Attackers used the exposed key to copy data from a read-only storage bucket Cursor used to share information with Fireworks. Although the bucket was read-only, the attackers could still copy its contents.

The bucket held historical AI coding-agent traces, mostly from 2025, from users who had not enabled Privacy Mode and whose data could be used for model training. Some traces contained API credentials that had not been redacted. The attackers used exposed credentials to access additional Cursor systems and steal data from several software-as-a-service tools.

Cursor says fewer than 0.05% of users had coding-agent traces exposed and fewer than 0.01% of customers had API credentials compromised. The company revoked the Fireworks key, rotated exposed credentials, and investigated further unauthorized access. It also moved inference traffic away from Fireworks and redeployed its production infrastructure as a precaution.

05

What kinds of Cursor data and credentials did attackers access?

Attackers used exposed credentials to steal data from several of Cursor’s software-as-a-service tools. The stolen material included archived, non-production Cursor source code repositories, some of which contained business contact information. The incident also reached customer-related information and summaries of some AI coding-agent conversations.

The accessed data included limited customer support information and email conversations, product usage metrics, customer contact details, and credentials customers had previously shared with Cursor’s support team. The credential chain began with data copied from a read-only bucket containing historical coding-agent traces. Some traces held unredacted API credentials that allowed access to additional Cursor systems.

Cursor reported that fewer than 0.05% of users had traces exposed and fewer than 0.01% of customers had API credentials compromised. It said customer code repositories, official application releases, production infrastructure, and code and prompts belonging to Privacy Mode users were not affected. Cursor revoked and rotated exposed credentials and investigated additional access.

06

What is an API credential, and why can stealing one give attackers access to other services?

In general, an API credential is a secret key or token that a software service uses to recognize and authorize a request from a user or program. It can let software connect to an application or data service without a person repeatedly signing in. Its exact permissions depend on the service and how the credential was set up.

That makes credentials sensitive: anyone who obtains a valid one may be able to make requests using the access it grants. In Cursor’s incident, historical coding-agent traces in a storage bucket contained some unredacted API credentials. Attackers used exposed credentials to reach additional Cursor systems and steal data from several software-as-a-service tools.

Cursor said fewer than 0.01% of customers had API credentials compromised. The company rotated exposed credentials and revoked the Fireworks AI access key. The article does not specify the permissions of each stolen credential or say that every credential opened every service. The incident illustrates why credentials should be protected and why leaked credentials need to be revoked or replaced quickly.

07

How do software extension marketplaces fit into the software supply chain, and why is control over publishing access important?

In general, a software extension marketplace is a channel where developers find and install add-ons for their tools. Because extensions add software to users’ working environments, the marketplace is part of how software reaches them. Publishing controls matter because they determine who can upload or change packages that users may trust and install.

In Cursor’s incident, attackers obtained a credential that allowed them to distribute extensions through its marketplace. On October 29, they used it to place malicious copies of Paradox’s Paradox Code extension in the cache serving Cursor’s VS Code extension marketplace. The copies attempted to install persistent SSH backdoors. They were available for about 11 hours.

The attack did not compromise Paradox’s legitimate extension or Open VSX, the marketplace Cursor uses. Cursor removed the builds, rebuilt its extension marketplace, rotated credentials, and redeployed production infrastructure as a precaution. The case shows why publishing credentials need protection: control of a distribution channel can let attackers send malicious software to users, even without compromising the original extension.

This brief was written by AI from the original reporting and checked by other models. Names, figures and quotes come from the source; read it for full context.

Read more in the JupiteX app

Pulse is free. New stories every 4 hours, each one broken into the questions that explain it.

Or read more news on the web