INSPECTING ARCHIVED INTELLIGENCE (OUTDATED VERSION).
Claude Code Extension Trojanized via Stolen Fireworks AI Cursor
|
2026-11-16 15:45
HIGH
LOW
AI-ENABLED ATTACK · AUTONOMOUS
DATA BREACH
SUPPLY CHAIN
MALWARE & BOTNETS
Executive Summary
AI-generated
A supply chain attack on Fireworks AI, which occurred in October 2026, led to a breach of credentials and the theft of customer data from Anthropic's Claude Code extension. The attackers used internal Fireworks credentials to access stored secrets and account information before compromising a third-party cloud provider, allowing them to reach development machines and steal customer data. A compromised Cursor access key was then used by the attacker to copy sensitive data from read-only storage buckets, including unredacted credentials that granted access to SaaS tools used by Cursor. The attack resulted in 5,000 users being affected by a tampered version of Anthropic's Claude Code extension on the Cursor VS Code Extension Marketplace for approximately 11 hours; malicious versions were made available through this marketplace.
Technical Mitigations AI-generated
• Follow the detection and remediation instructions in the official notice , remove the unauthorised SSH key and persistence files, and rotate accessible credentials from a trusted computer.
• Remove the malicious extension versions as soon as possible.
Technical Observables 2 IoCs Login Required
Indicators of Compromise (IoCs) Protected
Raw threat telemetry, malicious IP addresses, file hashes, and direct VirusTotal correlation are restricted to authenticated users.
ips 1
files 1
Intelligence Metadata
Actors / Malware / CVEs / Campaigns
Target & Sectors
Global Scope
technologytechnology
mediamedia
Incident Timeline
Oct. 1
Forever Security quietly patched the issue with its Cursor on October 1 after it failed to respond to a disclosure.
Click on any entity below to view its context and source!
infrastructure
Cursor
Cursor did not respond to Forever Security's disclosure, but the company quietly patched the issue on Oct. 1.
organisation
Forever Security's
Cursor did not respond to Forever Security's disclosure, but the company quietly patched the issue on Oct. 1.
October 29
Threat actors used a stolen credential to publish malicious copies of Anthropic's Claude Code extension on Cursor's VS Code extension marketplace.
Click on any entity below to view its context and source!
infrastructure
Cursor
The
supply chain attack
was discovered on October 29 and resulted in threat actors making two modified versions of the anthropic.claude-code extension (2.1.312 and 2.1.313) available through the Cursor VS Code Extension Marketplace for roughly 11 hours.
The breach was disclosed in a
new incident report
published Friday, which explains how attackers distributed
malicious versions of Anthropic's Claude Code extension
through Cursor's marketplace on October 29.
On October 29, the hackers used that credential to publish malicious copies of Anthropic's Claude Code extension on Cursor's VS Code extension marketplace.
infrastructure
Vs Code
The
supply chain attack
was discovered on October 29 and resulted in threat actors making two modified versions of the anthropic.claude-code extension (2.1.312 and 2.1.313) available through the Cursor VS Code Extension Marketplace for roughly 11 hours.
On October 29, the hackers used that credential to publish malicious copies of Anthropic's Claude Code extension on Cursor's VS Code extension marketplace.
"On October 29, using a stolen credential from the exposed data from our internal system, the attacker placed malicious, unauthorized copies of Anthropic's Claude Code extension directly on the cache used to serve the Cursor VS Code extension marketplace," Cursor explained.
infrastructure
2.1.313
The
supply chain attack
was discovered on October 29 and resulted in threat actors making two modified versions of the anthropic.claude-code extension (2.1.312 and 2.1.313) available through the Cursor VS Code Extension Marketplace for roughly 11 hours.
The incident occurred on October 29 and involved versions 2.1.312 and 2.1.313 of the
anthropic.claude-code
extension, which were live in the Cursor VS Code Extension Marketplace for roughly 11 hours.
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
infrastructure
2.1.312
The
supply chain attack
was discovered on October 29 and resulted in threat actors making two modified versions of the anthropic.claude-code extension (2.1.312 and 2.1.313) available through the Cursor VS Code Extension Marketplace for roughly 11 hours.
The incident occurred on October 29 and involved versions 2.1.312 and 2.1.313 of the
anthropic.claude-code
extension, which were live in the Cursor VS Code Extension Marketplace for roughly 11 hours.
general_metric
11 hours
The
supply chain attack
was discovered on October 29 and resulted in threat actors making two modified versions of the anthropic.claude-code extension (2.1.312 and 2.1.313) available through the Cursor VS Code Extension Marketplace for roughly 11 hours.
The incident occurred on October 29 and involved versions 2.1.312 and 2.1.313 of the
anthropic.claude-code
extension, which were live in the Cursor VS Code Extension Marketplace for roughly 11 hours.
victims
5,000 users
"At most 5,000 users who requested specific cached versions of the extension on October 29 potentially received the malicious build," Cursor said.
general_metric
66 line
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
tactic
T1059.007 - JavaScript
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
infrastructure
2.27
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
infrastructure
1.35
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
infrastructure
9.27
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
infrastructure
8.35
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
general_metric
2.27 am
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
general_metric
1.35 pm
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
general_metric
9.27 am
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
general_metric
8.35 pm
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
October 29, 2026
Threat actors used a stolen Fireworks AI key to target users who installed malicious versions of the Claude Code extension between October 29, 2026, and October 30, 2026.
Click on any entity below to view its context and source!
infrastructure
Cursor
The next phase of the attack took place on October 29, 2026, when the attacker used a stolen credential from the exposed data of its internal system to place trojanized versions of Anthropic's Claude Code extension directly on the cache used to serve Cursor's extension marketplace.
Cursor has revealed that the incident has affected at most 5,000 users who installed the malicious version between about 2:27 a.m. and 1:35 p.m. PT on October 29, 2026.
infrastructure
2.1.312
The company published its security notice on October 29, 2026, then updated it on October 30 to identify two affected versions, 2.1.312 and 2.1.313, for Windows and macOS.
infrastructure
2.1.313
The company published its security notice on October 29, 2026, then updated it on October 30 to identify two affected versions, 2.1.312 and 2.1.313, for Windows and macOS.
infrastructure
Macos
The company published its security notice on October 29, 2026, then updated it on October 30 to identify two affected versions, 2.1.312 and 2.1.313, for Windows and macOS.
infrastructure
Windows
The company published its security notice on October 29, 2026, then updated it on October 30 to identify two affected versions, 2.1.312 and 2.1.313, for Windows and macOS.
victims
5,000 users
Cursor has revealed that the incident has affected at most 5,000 users who installed the malicious version between about 2:27 a.m. and 1:35 p.m. PT on October 29, 2026.
2026/10/30
Weizman published research detailing weaknesses in top desktop AI assistants' browsers that could be exploited on the same day as the Anthropic Cursor attack.
Click on any entity below to view its context and source!
infrastructure
Cursor
More Security Concerns for Cursor, AI Assistants
On the same day of the Anthropic Cursor attack, Weizman published separate
research
detailing weaknesses in the browsers built into top desktop AI assistants today, which could be exploited to devastating effect.
November 1, 2026
Cursor detected no additional malicious activity since November 1, 2026.
Click on any entity below to view its context and source!
infrastructure
Cursor
Furthermore, Cursor emphasized that no additional malicious activity has been detected since November 1, 2026.
2026/11/08
Notion confirmed its data was stolen in the Fireworks AI hack, which also breached a stolen fireworks AI key used to target Claude Code Extension.
Click on any entity below to view its context and source!
organisation
Notion
Last week, online productivity startup Notion also confirmed its data was stolen in the Fireworks AI hack.
Nov 15, 2026
Ravie Lakshmanan Nov 15, 2026 Artificial Intelligence / Software Security The compromise of Anthropic's official Claude Code extension on Cursor's extension marketplace was the result of the theft of an access key that was exposed…
Click on any entity below to view its context and source!
infrastructure
Cursor
Ravie Lakshmanan
Nov 15, 2026
Artificial Intelligence / Software Security
The
compromise of Anthropic's official Claude Code extension
on Cursor's extension marketplace was the result of the theft of an access key that was exposed in the wake of a breach targeting
Fireworks AI
.
tactic
T1588.007 - Artificial Intelligence
Ravie Lakshmanan
Nov 15, 2026
Artificial Intelligence / Software Security
The
compromise of Anthropic's official Claude Code extension
on Cursor's extension marketplace was the result of the theft of an access key that was exposed in the wake of a breach targeting
Fireworks AI
.
2026/11/16
Threat actors compromised a Cursor access key issued to Fireworks AI, allowing them to copy data from a read-only storage bucket and subsequently use the credentials as a starting point to reach select software-as-a-service (SaaS) tools used by Cursor.
Click on any entity below to view its context and source!
infrastructure
Cursor
Stolen Fireworks AI Key Led to Cursor Data Breach and Trojanized Claude Code Extension.
Cursor Supply Chain Attack Started at Fireworks AI.
Credentials compromised in the hack of AI inference platform Fireworks AI allowed attackers to steal more secrets and push malicious Claude Code extension copies, Cursor says.
infrastructure
Vs Code
While Cursor blocked the attacker’s access immediately after discovering the unauthorized activity, the attacker used secrets exfiltrated from the company’s internal systems to place “malicious, unauthorized copies of Anthropic’s Claude Code extension directly on the cache used to serve the Cursor VS Code extension marketplace.”
“If customers installed or updated to the tampered version Anthropic.claude-code 2.1.312 or 2.1.313 from the Cursor VS Code extension marketplace during the exposure window, we ask customers to follow the steps in our
security notice
.
Anthropic's official Claude Code extension for the Cursor VS Code Marketplace was hijacked and swapped out for a malicious version with a persistent backdoor.
infrastructure
Macos
According to Cursor, because enabling the local SSH service requires admin rights and the extension does not have them on most macOS and Windows systems, the step failed, preventing the attackers from logging in through the tunnel.
Cursor
said coding errors and administrator-permission requirements blocked the SSH access step on most Windows and macOS computers.
Cursor provides separate cleanup scripts for Windows and macOS and advises assessing whether affected computers need to be rebuilt.
infrastructure
Windows
According to Cursor, because enabling the local SSH service requires admin rights and the extension does not have them on most macOS and Windows systems, the step failed, preventing the attackers from logging in through the tunnel.
Cursor
said coding errors and administrator-permission requirements blocked the SSH access step on most Windows and macOS computers.
Cursor provides separate cleanup scripts for Windows and macOS and advises assessing whether affected computers need to be rebuilt.
infrastructure
2.1.313
“If customers installed or updated to the tampered version Anthropic.claude-code 2.1.312 or 2.1.313 from the Cursor VS Code extension marketplace during the exposure window, we ask customers to follow the steps in our
security notice
.
Any software developers who downloaded or updated Claude Code in their
Cursor integrated development environment
(IDE) during that time would have received a malicious version 2.1.313, which pilfered sensitive data from infected machines and, in a minority of cases, could have allowed attackers persistent, privileged access to users' systems.
The malicious extensions, versions 2.1.312 and 2.1.313, were available for approximately 11 hours before Cursor removed them.
victims
5,000 users
According to Cursor, at most 5,000 users were affected.
Cursor
stated
that "at most 5,000 users were affected," and that they will be contacted about the incident directly.
Cursor estimates that up to 5,000 users who requested affected extension versions during the exposure window may have received the malicious builds.
infrastructure
2.1.312
“If customers installed or updated to the tampered version Anthropic.claude-code 2.1.312 or 2.1.313 from the Cursor VS Code extension marketplace during the exposure window, we ask customers to follow the steps in our
security notice
.
The malicious extensions, versions 2.1.312 and 2.1.313, were available for approximately 11 hours before Cursor removed them.
"
The company said it notified affected users who installed versions 2.1.312 or 2.1.313 from the marketplace during the exposure window and helped revoke exposed credentials.
organisation
Microsoft
Cursor is an AI-powered code editor built on Microsoft's open-source Visual Studio Code (VS Code), allowing developers to write, modify, and debug software using AI assistants.
The malicious update was distributed only through Cursor's marketplace, with no evidence that it reached Open VSX or Microsoft's official VS Code Marketplace.
The modified extension versions were not developed by Anthropic, and neither was pushed to the Microsoft Visual Studio Marketplace, Open VSX, or the NPM registry.
organisation
Fireworks AI
According to Cursor, the attack began when hackers obtained an access key stored on Fireworks AI, a third-party provider used for AI model training and inference.
"
Although the name of the provider was not disclosed, TechCrunch
reported
that the provider in question is Fireworks AI.
organisation
Limited
The attacker, Cursor added, used these credentials as a starting point to reach select software-as-a-service (SaaS) tools used by Cursor and then access the following data -
Select archived non-production Cursor source code repositories, including some containing business contact information
Limited customer support data and email conversations
Product usage metrics and contact details…
Limited customer support information and email conversations.
organisation
SSH
Cybersecurity researchers have
disclosed
that a trojanized version of Anthropic's official Claude Code extension was distributed through the Cursor Marketplace as a legitimate update, attempting to install a persistent SSH backdoor on affected systems.
It attempts to add an attacker-controlled key to victims' SSH configurations, with the goal of establishing reverse tunnels that act as a
backdoor
into the compromised systems.
organisation
LaunchAgent
The malicious extension also attempts to make those connections persistent by registering background processes — for macOS systems, a LaunchAgent; and for Windows, scheduled tasks running with system privileges, plus a fallback Startup script.
Subsequently, it attempts to make the victim reachable through SSH across Windows, Linux, and macOS, establishes a reverse tunnel to "
134.209.82[.]225
," installs platform-specific persistence in the form of a scheduled task, crontab, and LaunchAgent, and transmits host reconnaissance data to "134.209.82[.]225:8099/beacon" on port 8099 over an unencrypted HTTP request.
financial
$5 Million
Related:
Maersk Confirms Data Theft Amid $5 Million Extortion Attempt
Related:
Revolve Puts Data Breach Victim Count at 1.5 Million
Related:
Iris Plaza Breach May Have Exposed Data of 7.7 Million Customers
Related:
Over 3 Million People Hit by Frontline Education Data Breach
organisation
Cursor Data Breach
Stolen Fireworks AI Key Led to Cursor Data Breach and Trojanized Claude Code Extension.
organisation
Dark Reading
Related:
Anthropic-Themed Phishing Emails Bypass Top Security Tools
Cursor did not immediately respond to an inquiry from Dark Reading regarding both of these stories.
organisation
IDE
Any software developers who downloaded or updated Claude Code in their
Cursor integrated development environment
(IDE) during that time would have received a malicious version 2.1.313, which pilfered sensitive data from infected machines and, in a minority of cases, could have allowed attackers persistent, privileged access to users' systems.
organisation
Claude Desktop
His test cases included Claude Desktop, GitHub Copilot, Codex, VS Code, and Cursor.
organisation
GitHub Copilot
His test cases included Claude Desktop, GitHub Copilot, Codex, VS Code, and Cursor.
infrastructure
Visual Studio Code
Cursor is an AI-powered code editor built on Microsoft's open-source Visual Studio Code (VS Code), allowing developers to write, modify, and debug software using AI assistants.
organisation
Anthropic’s Claude Code
Cursor users have been targeted with malicious versions of Anthropic’s Claude Code extension in a new supply chain attack.
organisation
Cursor Marketplace
The disclosure comes nearly two weeks after a
trojanized version
of Anthropic's official Claude Code extension was distributed through Cursor Marketplace as a legitimate update, attempting to install a persistent SSH backdoor on affected systems.
organisation
Select
The attacker, Cursor added, used these credentials as a starting point to reach select software-as-a-service (SaaS) tools used by Cursor and then access the following data -
Select archived non-production Cursor source code repositories, including some containing business contact information
Limited customer support data and email conversations
Product usage metrics and contact details…
organisation
Persistent SSH Backdoor
Official Claude Code Extension on Cursor Marketplace Trojanized with Persistent SSH Backdoor.
organisation
the Cursor Marketplace
Cybersecurity researchers have
disclosed
that a trojanized version of Anthropic's official Claude Code extension was distributed through the Cursor Marketplace as a legitimate update, attempting to install a persistent SSH backdoor on affected systems.
organisation
The Hacker News
"The package identifies itself as anthropic.claude-code version 2.1.313, but its main JavaScript bundle contains a malicious 66-line loader inserted before the legitimate extension code," Jenn Gile, co-founder of OpenSourceMalware, said in a statement shared with The Hacker News.
organisation
Adaptive Security's
Adaptive Security's AI agents analyze email intent, context, and behavior to uncover phishing, business email compromise, and invoice fraud.
infrastructure
Linux
Subsequently, it attempts to make the victim reachable through SSH across Windows, Linux, and macOS, establishes a reverse tunnel to "
134.209.82[.]225
," installs platform-specific persistence in the form of a scheduled task, crontab, and LaunchAgent, and transmits host reconnaissance data to "134.209.82[.]225:8099/beacon" on port 8099 over an unencrypted HTTP request.
organisation
Forever Security
"IDE extensions are scarier than browser extensions because, unlike browser extensions, they are designed to have the same level of permissions as their host," Gal Weizman, security researcher at Forever Security, tells Dark Reading.
organisation
Claude
As soon as the extension loaded — without need for the user to establish a Claude conversation — the new file is revved up.
organisation
IP
Stolen data was limited to hostnames, usernames, operating systems (OS), OS architectures, home directory paths, environment variable names, working directories, configured Git emails, and public IP addresses.
Tactical Metrics
Metrics
financial
$5M
Ransom
Click for context!
Related:
Maersk Confirms Data Theft Amid $5 Million Extortion Attempt
Related:
Revolve Puts Data Breach Victim Count at 1.5 Million
Related:
Iris Plaza Breach May Have Exposed Data of 7.7 Million Customers
Related:
Over 3 Million People Hit by Frontline Education Data Breach
Metrics
infrastructure
Cursor
Affected Product
Cursor Supply Chain Attack Started at Fireworks AI.
Credentials compromised in the hack of AI inference platform Fireworks AI allowed attackers to steal more secrets and push malicious Claude Code extension copies, Cursor says.
The
supply chain attack
was discovered on October 29 and resulted in threat actors making two modified versions of the anthropic.claude-code extension (2.1.312 and 2.1.313) available through the Cursor VS Code Extension Marketplace for roughly 11 hours.
Anthropic was never hacked, and all the malicious activity occurred in Cursor’s environment, the company explains.
Instead, Cursor was hit after Fireworks was
hacked twice
in October.
A Cursor access key issued to Fireworks was compromised in one of these incidents and the attacker used it to copy data from a read-only storage bucket, including unredacted credentials that opened the door to SaaS tools used by Cursor.
“From these tools they accessed select archived non-production Cursor source code repositories (some containing business contact information), limited customer support data and email conversations, product usage metrics and contact details, summaries of some coding agent conversations, and some credentials users had sent to the Cursor support team,”
Cursor explains
.
While Cursor blocked the attacker’s access immediately after discovering the unauthorized activity, the attacker used secrets exfiltrated from the company’s internal systems to place “malicious, unauthorized copies of Anthropic’s Claude Code extension directly on the cache used to serve the Cursor VS Code extension marketplace.”
According to Cursor, at most 5,000 users were affected.
“If customers installed or updated to the tampered version Anthropic.claude-code 2.1.312 or 2.1.313 from the Cursor VS Code extension marketplace during the exposure window, we ask customers to follow the steps in our
security notice
.
Uninstalling the extension alone may not be enough,” Cursor notes.
Related:
Anthropic-Themed Phishing Emails Bypass Top Security Tools
Cursor did not immediately respond to an inquiry from Dark Reading regarding both of these stories.
Devs Hit by Trojanized Claude Code Extension for Cursor.
Anthropic's official Claude Code extension for the Cursor VS Code Marketplace was hijacked and swapped out for a malicious version with a persistent backdoor.
The
SolarWinds
-style attack in miniature lasted only 11 hours, from 9:27 a.m. GMT to 8:35 p.m. GMT, on Oct. 29, according to Cursor's disclosure.
Any software developers who downloaded or updated Claude Code in their
Cursor integrated development environment
(IDE) during that time would have received a malicious version 2.1.313, which pilfered sensitive data from infected machines and, in a minority of cases, could have allowed attackers persistent, privileged access to users' systems.
Cursor
stated
that "at most 5,000 users were affected," and that they will be contacted about the incident directly.
Separately that same day, researchers described an unrelated exploit in Cursor, allowing attackers to deploy arbitrary code to developers' machines via Cursor's built-in browser.
Cursor's Claude Code Conundrum
For some, the notion of a compromised "extension" might evoke browser extensions, which are so frequently vulnerable, suspicious, and outright malicious that cybersecurity news concerning them tends not to evoke much panic, even in
extreme cases
.
This whole business was doomed from the start, Cursor wrote, since turning on the local SSH service requires administrative privileges, which the
Claude Code
extension only possesses in "highly unusual circumstances."
Related:
'Gheist' Weaponizes GitHub Tokens to Steal Cloud Secrets
Victims weren't totally out of the woods, though, as the malicious code also exfiltrated system information via…
Cursor Contains Compromise
Cursor's public disclosure suggests that the security incident was its own, not Anthropic's.
In response to the discovery, Cursor took a variety of steps to protect itself: rotating its keys, checking for tampering in other extensions, and taking its marketplace temporarily offline and rebuilding it.
Cursor did not specify how the intrusion occurred in the first place.
More Security Concerns for Cursor, AI Assistants
On the same day of the Anthropic Cursor attack, Weizman published separate
research
detailing weaknesses in the browsers built into top desktop AI assistants today, which could be exploited to devastating effect.
His test cases included Claude Desktop, GitHub Copilot, Codex, VS Code, and Cursor.
Cursor has arguably the worst issue of the bunch.
Then they opened up the site in a built-in Cursor browser.
"When the command hit Cursor's safety sandbox and was blocked, the agent — on its own, with nothing in the page telling it to — switched the sandbox off and tried again.
Cursor did not respond to Forever Security's disclosure, but the company quietly patched the issue on Oct. 1.
"Here we've had two examples of that: built-in browsers, and what happened with Claude Code, both of which impacted Cursor," he concludes.
Cursor confirms security breach led to malicious Claude Code extension.
Cursor confirmed it suffered a security breach that let hackers distribute a malicious Claude Code extension through its marketplace after stealing credentials from AI infrastructure provider Fireworks AI.
The breach was disclosed in a
new incident report
published Friday, which explains how attackers distributed
malicious versions of Anthropic's Claude Code extension
through Cursor's marketplace on October 29.
Cursor is an AI-powered code editor built on Microsoft's open-source Visual Studio Code (VS Code), allowing developers to write, modify, and debug software using AI assistants.
According to Cursor, the attack began when hackers obtained an access key stored on Fireworks AI, a third-party provider used for AI model training and inference.
The attackers used the exposed key to copy data from a read-only storage bucket Cursor used to share information with Fireworks.
Some of these traces contained unredacted API credentials, allowing the attackers to access additional Cursor systems.
Cursor says fewer than 0.05% of users had coding-agent traces exposed, while fewer than 0.01% of customers had API credentials compromised.
Using the exposed credentials, the hackers stole data from several of Cursor's software-as-a-service (SaaS) tools, including:
Archived, non-production Cursor source code repositories, some containing business contact information.
Credentials that customers had previously shared with Cursor's support team.
Cursor says its monitoring systems detected the unauthorized activity, leading the company to revoke the Fireworks access key, rotate exposed credentials, and investigate additional unauthorized access.
As the investigation progressed, Cursor temporarily turned off some internal and product services as a precaution.
However, the attackers had already obtained a credential that allowed them to distribute extensions through Cursor's marketplace.
On October 29, the hackers used that credential to publish malicious copies of Anthropic's Claude Code extension on Cursor's VS Code extension marketplace.
"On October 29, using a stolen credential from the exposed data from our internal system, the attacker placed malicious, unauthorized copies of Anthropic's Claude Code extension directly on the cache used to serve the Cursor VS Code extension marketplace," Cursor explained.
The attack did not compromise Anthropic's legitimate extension or Open VSX, the extension marketplace Cursor uses.
The malicious extensions, versions 2.1.312 and 2.1.313, were available for approximately 11 hours before Cursor removed them.
Cursor estimates that up to 5,000 users who requested affected extension versions during the exposure window may have received the malicious builds.
Cursor says it has moved all inference traffic away from Fireworks, revoked and rotated exposed credentials, rebuilt its extension marketplace, and redeployed its production infrastructure as a precaution.
Customers who installed or updated to the affected Claude Code extension during the attack are advised to follow
Cursor's remediation guidance
.
Malicious Claude Code Extension Served in Cursor Supply Chain Attack.
Cursor users have been targeted with malicious versions of Anthropic’s Claude Code extension in a new supply chain attack.
The incident occurred on October 29 and involved versions 2.1.312 and 2.1.313 of the
anthropic.claude-code
extension, which were live in the Cursor VS Code Extension Marketplace for roughly 11 hours.
“A malicious actor used access to Cursor’s marketplace to publish a tampered copy of the Claude Code extension, which we didn’t build or publish.
The versions on our official channels are not affected, and this only impacts Cursor users,” Anthropic’s Felix Rieseberg
said
.
Cursor removed the modified Claude Code extension from the store immediately after identifying it and suspended the marketplace to investigate the incident.
Cursor found no other malicious extensions, rotated all keys, and rebuilt and reopened the marketplace.
“We have no evidence that other versions of the extension or the official distribution are affected,”
Cursor said
.
According to Cursor, because enabling the local SSH service requires admin rights and the extension does not have them on most macOS and Windows systems, the step failed, preventing the attackers from logging in through the tunnel.
Cursor has released a script to help users detect infections, along with indicators of compromise (IoCs) to identify potential compromises across environments.
Uninstalling the extension may not be sufficient — the payload drops standalone persistence scripts that keep running,” Cursor notes.
Stolen Fireworks AI Key Led to Cursor Data Breach and Trojanized Claude Code Extension.
Ravie Lakshmanan
Nov 15, 2026
Artificial Intelligence / Software Security
The
compromise of Anthropic's official Claude Code extension
on Cursor's extension marketplace was the result of the theft of an access key that was exposed in the wake of a breach targeting
Fireworks AI
.
"An attacker obtained an access key we had issued to Fireworks," Cursor
said
in a Friday update.
The disclosure comes nearly two weeks after a
trojanized version
of Anthropic's official Claude Code extension was distributed through Cursor Marketplace as a legitimate update, attempting to install a persistent SSH backdoor on affected systems.
Privacy Mode is a
Cursor-specific setting
that, if enabled, allows users to opt out of their data being used for training purposes.
The attacker, Cursor added, used these credentials as a starting point to reach select software-as-a-service (SaaS) tools used by Cursor and then access the following data -
Select archived non-production Cursor source code repositories, including some containing business contact information
Limited customer support data and email conversations
Product usage metrics and contact details…
The company also said it temporarily disabled internal and product services, disrupting access to some Cursor features.
The next phase of the attack took place on October 29, 2026, when the attacker used a stolen credential from the exposed data of its internal system to place trojanized versions of Anthropic's Claude Code extension directly on the cache used to serve Cursor's extension marketplace.
"At most 5,000 users who requested specific cached versions of the extension on October 29 potentially received the malicious build," Cursor said.
Furthermore, Cursor emphasized that no additional malicious activity has been detected since November 1, 2026.
There is no evidence that the incident impacted users who had enabled Privacy Mode, customer code repositories, Cursor's own production environment, Cursor app and its official updates, its other apps, or any other extension or plugin downloaded from Cursor Marketplace.
Cursor Distributed Claude Code Extensions With SSH Backdoor Code.
Developers installing
Claude Code
through Cursor’s extension marketplace were offered tampered packages designed to collect information about their computers and give attackers remote access.
Cursor removed the malicious downloads but warned that deleting the extension alone will not remove the files it left behind.
Cursor
said coding errors and administrator-permission requirements blocked the SSH access step on most Windows and macOS computers.
(Source: Microsoft)
Removing the Extension Is Only Part of the Cleanup
Cursor temporarily closed the marketplace, rotated keys and rebuilt it before restoring service.
Cursor provides separate cleanup scripts for Windows and macOS and advises assessing whether affected computers need to be rebuilt.
Editor’s note:
Cursor’s initial notice said at most 5,000 users were affected.
Cursor’s original notice stated that at most 5,000 users were affected, a figure removed from the updated version.
Official Claude Code Extension on Cursor Marketplace Trojanized with Persistent SSH Backdoor.
Cybersecurity researchers have
disclosed
that a trojanized version of Anthropic's official Claude Code extension was distributed through the Cursor Marketplace as a legitimate update, attempting to install a persistent SSH backdoor on affected systems.
The malicious update, version 2.1.313, was distributed under the extension's legitimate publisher identity, prompting Cursor to temporarily shut down its marketplace to remove the compromised package.
"We identified a security incident affecting a modified Claude Code VS Code extension accessed through our Cursor IDE extension marketplace, impacting at most 5,000 users," Cursor
said
.
It does not impact other extensions downloaded from the cursor[.]com/marketplace, Grok Bot, Grok, or our other SpaceXSI product.
However, Cursor's statement that the malicious version "was not developed by Anthropic" misses the point.
The malicious update was distributed only through Cursor's marketplace, with no evidence that it reached Open VSX or Microsoft's official VS Code Marketplace.
Cursor has revealed that the incident has affected at most 5,000 users who installed the malicious version between about 2:27 a.m. and 1:35 p.m. PT on October 29, 2026.
The incident has since been traced back to a compromise of the Cursor VS Code Marketplace.)
Metrics
infrastructure
Vs Code
Affected Product
The
supply chain attack
was discovered on October 29 and resulted in threat actors making two modified versions of the anthropic.claude-code extension (2.1.312 and 2.1.313) available through the Cursor VS Code Extension Marketplace for roughly 11 hours.
While Cursor blocked the attacker’s access immediately after discovering the unauthorized activity, the attacker used secrets exfiltrated from the company’s internal systems to place “malicious, unauthorized copies of Anthropic’s Claude Code extension directly on the cache used to serve the Cursor VS Code extension marketplace.”
“If customers installed or updated to the tampered version Anthropic.claude-code 2.1.312 or 2.1.313 from the Cursor VS Code extension marketplace during the exposure window, we ask customers to follow the steps in our
security notice
.
Anthropic's official Claude Code extension for the Cursor VS Code Marketplace was hijacked and swapped out for a malicious version with a persistent backdoor.
His test cases included Claude Desktop, GitHub Copilot, Codex, VS Code, and Cursor.
Cursor is an AI-powered code editor built on Microsoft's open-source Visual Studio Code (VS Code), allowing developers to write, modify, and debug software using AI assistants.
On October 29, the hackers used that credential to publish malicious copies of Anthropic's Claude Code extension on Cursor's VS Code extension marketplace.
"On October 29, using a stolen credential from the exposed data from our internal system, the attacker placed malicious, unauthorized copies of Anthropic's Claude Code extension directly on the cache used to serve the Cursor VS Code extension marketplace," Cursor explained.
The incident occurred on October 29 and involved versions 2.1.312 and 2.1.313 of the
anthropic.claude-code
extension, which were live in the Cursor VS Code Extension Marketplace for roughly 11 hours.
"We identified a security incident affecting a modified Claude Code VS Code extension accessed through our Cursor IDE extension marketplace, impacting at most 5,000 users," Cursor
said
.
The malicious update was distributed only through Cursor's marketplace, with no evidence that it reached Open VSX or Microsoft's official VS Code Marketplace.
As of October 30, 2026, the malicious version has been removed, and Anthropic has
published version 2.1.314
to both Open VSX and Microsoft VS Code Marketplace.
The extension is designed to activate after VS Code startup and when its webview opens, then proceeds to add an attacker-controlled key to the user's "~/.ssh/authorized_keys" file.
The incident has since been traced back to a compromise of the Cursor VS Code Marketplace.)
Metrics
infrastructure
2.1.312
Software Version
The
supply chain attack
was discovered on October 29 and resulted in threat actors making two modified versions of the anthropic.claude-code extension (2.1.312 and 2.1.313) available through the Cursor VS Code Extension Marketplace for roughly 11 hours.
“If customers installed or updated to the tampered version Anthropic.claude-code 2.1.312 or 2.1.313 from the Cursor VS Code extension marketplace during the exposure window, we ask customers to follow the steps in our
security notice
.
The malicious extensions, versions 2.1.312 and 2.1.313, were available for approximately 11 hours before Cursor removed them.
The incident occurred on October 29 and involved versions 2.1.312 and 2.1.313 of the
anthropic.claude-code
extension, which were live in the Cursor VS Code Extension Marketplace for roughly 11 hours.
"
The company said it notified affected users who installed versions 2.1.312 or 2.1.313 from the marketplace during the exposure window and helped revoke exposed credentials.
The company published its security notice on October 29, 2026, then updated it on October 30 to identify two affected versions, 2.1.312 and 2.1.313, for Windows and macOS.
That figure has since been removed, and the updated notice identifies version 2.1.312 alongside 2.1.313.
Metrics
infrastructure
2.1.313
Software Version
The
supply chain attack
was discovered on October 29 and resulted in threat actors making two modified versions of the anthropic.claude-code extension (2.1.312 and 2.1.313) available through the Cursor VS Code Extension Marketplace for roughly 11 hours.
“If customers installed or updated to the tampered version Anthropic.claude-code 2.1.312 or 2.1.313 from the Cursor VS Code extension marketplace during the exposure window, we ask customers to follow the steps in our
security notice
.
Any software developers who downloaded or updated Claude Code in their
Cursor integrated development environment
(IDE) during that time would have received a malicious version 2.1.313, which pilfered sensitive data from infected machines and, in a minority of cases, could have allowed attackers persistent, privileged access to users' systems.
The malicious extensions, versions 2.1.312 and 2.1.313, were available for approximately 11 hours before Cursor removed them.
The incident occurred on October 29 and involved versions 2.1.312 and 2.1.313 of the
anthropic.claude-code
extension, which were live in the Cursor VS Code Extension Marketplace for roughly 11 hours.
"
The company said it notified affected users who installed versions 2.1.312 or 2.1.313 from the marketplace during the exposure window and helped revoke exposed credentials.
The company published its security notice on October 29, 2026, then updated it on October 30 to identify two affected versions, 2.1.312 and 2.1.313, for Windows and macOS.
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
That figure has since been removed, and the updated notice identifies version 2.1.312 alongside 2.1.313.
The malicious update, version 2.1.313, was distributed under the extension's legitimate publisher identity, prompting Cursor to temporarily shut down its marketplace to remove the compromised package.
The attackers distributed it as version 2.1.313 under Anthropic's legitimate Claude Code extension ID (anthropic.claude-code), rather than as a separate fake extension.
"The package identifies itself as anthropic.claude-code version 2.1.313, but its main JavaScript bundle contains a malicious 66-line loader inserted before the legitimate extension code," Jenn Gile, co-founder of OpenSourceMalware, said in a statement shared with The Hacker News.
Metrics
victims
5,000
Users
According to Cursor, at most 5,000 users were affected.
Cursor
stated
that "at most 5,000 users were affected," and that they will be contacted about the incident directly.
Cursor estimates that up to 5,000 users who requested affected extension versions during the exposure window may have received the malicious builds.
"At most 5,000 users who requested specific cached versions of the extension on October 29 potentially received the malicious build," Cursor said.
Editor’s note:
Cursor’s initial notice said at most 5,000 users were affected.
Cursor’s original notice stated that at most 5,000 users were affected, a figure removed from the updated version.
"We identified a security incident affecting a modified Claude Code VS Code extension accessed through our Cursor IDE extension marketplace, impacting at most 5,000 users," Cursor
said
.
Cursor has revealed that the incident has affected at most 5,000 users who installed the malicious version between about 2:27 a.m. and 1:35 p.m. PT on October 29, 2026.
Metrics
infrastructure
Macos
Affected Product
The malicious extension also attempts to make those connections persistent by registering background processes — for macOS systems, a LaunchAgent; and for Windows, scheduled tasks running with system privileges, plus a fallback Startup script.
However, security researchers found coding errors that likely prevented the backdoor from working on many Windows and macOS systems where SSH was not already enabled.
Aiming at both macOS and Windows users, the extension contained JavaScript code designed to run when modules were loaded to establish a reverse SSH tunnel.
According to Cursor, because enabling the local SSH service requires admin rights and the extension does not have them on most macOS and Windows systems, the step failed, preventing the attackers from logging in through the tunnel.
The company published its security notice on October 29, 2026, then updated it on October 30 to identify two affected versions, 2.1.312 and 2.1.313, for Windows and macOS.
Cursor
said coding errors and administrator-permission requirements blocked the SSH access step on most Windows and macOS computers.
These included a macOS LaunchAgent, a Windows scheduled task named
SysMon
and a fallback script in the Windows Startup folder.
Cursor provides separate cleanup scripts for Windows and macOS and advises assessing whether affected computers need to be rebuilt.
Subsequently, it attempts to make the victim reachable through SSH across Windows, Linux, and macOS, establishes a reverse tunnel to "
134.209.82[.]225
," installs platform-specific persistence in the form of a scheduled task, crontab, and LaunchAgent, and transmits host reconnaissance data to "134.209.82[.]225:8099/beacon" on port 8099 over an unencrypted HTTP request.
"Turning on the local SSH service requires admin rights, which the malicious extension does not have on most macOS and Windows hosts," the SpaceXAI-owned company
said
in its own analysis of the incident.
Metrics
infrastructure
Windows
Affected Product
The malicious extension also attempts to make those connections persistent by registering background processes — for macOS systems, a LaunchAgent; and for Windows, scheduled tasks running with system privileges, plus a fallback Startup script.
However, security researchers found coding errors that likely prevented the backdoor from working on many Windows and macOS systems where SSH was not already enabled.
Aiming at both macOS and Windows users, the extension contained JavaScript code designed to run when modules were loaded to establish a reverse SSH tunnel.
According to Cursor, because enabling the local SSH service requires admin rights and the extension does not have them on most macOS and Windows systems, the step failed, preventing the attackers from logging in through the tunnel.
The company published its security notice on October 29, 2026, then updated it on October 30 to identify two affected versions, 2.1.312 and 2.1.313, for Windows and macOS.
Cursor
said coding errors and administrator-permission requirements blocked the SSH access step on most Windows and macOS computers.
These included a macOS LaunchAgent, a Windows scheduled task named
SysMon
and a fallback script in the Windows Startup folder.
Cursor provides separate cleanup scripts for Windows and macOS and advises assessing whether affected computers need to be rebuilt.
Subsequently, it attempts to make the victim reachable through SSH across Windows, Linux, and macOS, establishes a reverse tunnel to "
134.209.82[.]225
," installs platform-specific persistence in the form of a scheduled task, crontab, and LaunchAgent, and transmits host reconnaissance data to "134.209.82[.]225:8099/beacon" on port 8099 over an unencrypted HTTP request.
"Turning on the local SSH service requires admin rights, which the malicious extension does not have on most macOS and Windows hosts," the SpaceXAI-owned company
said
in its own analysis of the incident.
Metrics
infrastructure
Visual Studio Code
Affected Product
Cursor is an AI-powered code editor built on Microsoft's open-source Visual Studio Code (VS Code), allowing developers to write, modify, and debug software using AI assistants.
Metrics
infrastructure
Microsoft 365
Affected Product
See what's getting through with a 14-day read-only scan of Microsoft 365 or Google Workspace, with no MX changes.
Metrics
financial
$740K
Fraud Losses
At Railway, attackers stole customer authentication secrets, which were later used to generate approximately $740,000 in fraudulent AI charges.
Metrics
infrastructure
2.1.314
Software Version
To avoid confusion, Anthropic updated its legitimate Claude Code extension directly to version 2.1.314.
As of October 30, 2026, the malicious version has been removed, and Anthropic has
published version 2.1.314
to both Open VSX and Microsoft VS Code Marketplace.
Metrics
infrastructure
2.27
Software Version
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
Metrics
infrastructure
1.35
Software Version
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
Metrics
infrastructure
9.27
Software Version
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
Metrics
infrastructure
8.35
Software Version
The packages were available between approximately 2.27 am and 1.35 pm Pacific Daylight Time on October 29, equivalent to 9.27 am to 8.35 pm GMT
Malicious Code Ran Before the Extension Started
OpenSourceMalware’s
analysis
of version 2.1.313 found a 66-line malicious loader inserted before the extension’s normal JavaScript code.
Metrics
infrastructure
Linux
Affected Product
Subsequently, it attempts to make the victim reachable through SSH across Windows, Linux, and macOS, establishes a reverse tunnel to "
134.209.82[.]225
," installs platform-specific persistence in the form of a scheduled task, crontab, and LaunchAgent, and transmits host reconnaissance data to "134.209.82[.]225:8099/beacon" on port 8099 over an unencrypted HTTP request.
Intelligence Sources
The Hacker News
2026-10-30
Dark Reading
2026-10-30
SecurityWeek
2026-11-16
Cursor Supply Chain Attack Started at Fireworks AI
SecurityWeek
HackRead
2026-11-03
The Hacker News
2026-11-15
BleepingComputer
2026-11-14
SecurityWeek
2026-11-02
Unpublish from Social Media?
Are you sure you want to delete this podcast video from all synchronized social networks (YouTube, Facebook, Threads)?
Important:
Due to Meta API restrictions, Instagram Reels cannot be deleted automatically via API by third-party apps.
View Profile to Delete Manually
View Profile to Delete Manually
Tactical Intelligence
Report Intelligence Issue
Podcast Options
Generate
Incident Version History
CURRENT VERSION
Last Updated: 2026-11-17T07:05
Comprehensive Tactical Telemetry
Highly Correlated Entities
45x
organisation
Identified Entity
Dark Reading
entity
16x
timeline
Temporal Reference
October 29
date
7x
infrastructure
Affected Product
Cursor
software
7x
infrastructure
Software Version
2.1.312
version
4x
tactic
Cyber Operation Type
Extortion
tactic
4x
tactic
MITRE ATT&CK Technique
T1589.001 - Credentials
technique
2x
general metric
%
0
%
2x
general metric
Am
2
am
2x
general metric
Pm
1
pm
Contextual Telemetry
Context Block
10 METRICS
financial
Ransom
$5M
general metric
People
3,000,000
people
general metric
Hours
11
hours
victims
Users
5,000
users
general metric
Security Practices
10
security practices
general metric
Line
66
line
financial
Fraud Losses
$740K
general metric
Developers
5,000
developers
general metric
Environment Variables
40
environment variables
malware
Malware Payload
GlassWorm
tool
Click on any entity below to view its context in the main text!
Selective Unpublish
Select the networks you want to remove this post from. The system will try to delete the real post through the API and clean the database so you can publish it again.
By navigating this website, you accept the use of strictly necessary technical cookies for session security and basic platform functionality. We do not use tracking or advertising cookies. Read our Privacy Policy.
Generating report…
0s