Skip to content
← All stories

Application Security

Cursor breach let hackers spread malware through code extension

Hackers broke into Cursor and used stolen credentials to distribute a malicious Claude Code extension, exposing up to 5,000 developers to malware and leaking some customer data.

Ross Grant3 min read

Personal records spilling out of an open database

The short version

Hackers broke into Cursor, an AI code editor, by stealing credentials from a third-party provider. They used this access to upload a malicious code extension that could install malware on developers' computers. Up to 5,000 users may have been exposed before the extension was removed. Cursor has taken steps to fix the problem and is advising affected users on what to do next.

By the numbers

  • 5,000developers potentially exposed to malwareBleepingComputer
  • 0.05%users with coding-agent traces exposedBleepingComputer
  • 0.01%customers with API credentials compromisedBleepingComputer
  • 11 hoursmalicious extension was available on the marketplaceBleepingComputer
  • $740,000fraudulent AI charges at Railway (related incident)BleepingComputer

What happened?

Cursor confirmed a security breach that let hackers distribute a malicious version of the Claude Code extension through its marketplace after stealing credentials from Fireworks AI. The attack exposed up to 5,000 developers to malware and led to the theft of some customer data. Cursor has taken steps to contain the breach and advises affected users to follow its remediation guidance. Only BleepingComputer has reported on this so far.

What was taken

Attackers accessed internal Cursor systems and stole customer data, including historical AI coding-agent traces. Some of these traces contained unredacted API credentials. Using these, the attackers accessed additional Cursor services and stole:

  • Archived, non-production source code repositories (some with business contact info)
  • Limited customer support information and email conversations
  • Product usage metrics and customer contact details
  • Summaries of some AI coding-agent conversations
  • Credentials customers had shared with Cursor's support team

Cursor says fewer than 0.05% of users had coding-agent traces exposed, and fewer than 0.01% had API credentials compromised. The company states that production infrastructure, official application releases, customer code repositories, and data from users with Privacy Mode enabled were not affected.

How the attack worked

The breach began when attackers stole an access key from Fireworks AI, a third-party provider Cursor used for AI model training. The attackers used this key to copy data from a storage bucket, which included coding-agent traces with unredacted credentials. These credentials let them access more Cursor systems and eventually obtain a credential that allowed them to distribute extensions through Cursor's marketplace.

On 29 October, the attackers published malicious versions (2.1.312 and 2.1.313) of Anthropic's Claude Code extension on the Cursor VS Code extension marketplace. The malware collected system information, installed an attacker-controlled SSH key, and tried to enable persistent remote access. Coding errors in the malware likely prevented it from working on many Windows and macOS systems where SSH was not already enabled.

The malicious extensions were available for about 11 hours before Cursor removed them. Up to 5,000 users who requested the affected versions during that time may have received the malware.

The response

Cursor detected the unauthorised activity with its monitoring systems and took several steps:

  • Revoked the Fireworks access key
  • Rotated exposed credentials
  • Investigated further unauthorised access
  • Temporarily disabled some internal and product services
  • Rebuilt its extension marketplace
  • Redeployed production infrastructure
  • Moved all inference traffic away from Fireworks

Customers who installed or updated to the affected Claude Code extension during the attack are advised to follow Cursor's remediation guidance.

Has this happened before?

This incident is part of a wider trend. Security firm Wiz has called it "Black October" due to a wave of attacks on AI developer platforms. Other companies hit include Railway, Northflank, Lightning AI, OpenCode, and Warp. At Railway, attackers used stolen secrets to generate $740,000 in fraudulent AI charges. Fireworks AI, the third-party provider involved in the Cursor breach, also reported multiple security incidents in October, including exposed API keys and tokens, and unauthorised access to development machines.

What you should do

Developers who installed the affected Claude Code extension versions should check for signs of compromise, such as unexpected SSH keys or suspicious system activity, and follow Cursor's remediation steps. All users should review their credential management and consider enabling privacy features that limit data exposure.

Source: BleepingComputer (opens in a new tab)

Why it matters

This breach shows how attacks on software supply chains can quickly spread malware to thousands of developers. It also highlights the risks of storing sensitive credentials in AI training data and the importance of strong third-party security.

Key points

  1. Hackers used stolen credentials to upload malware to Cursor's extension marketplace.
  2. Up to 5,000 developers may have installed the malicious Claude Code extension.
  3. Some customer data, including API credentials, was also stolen.
  4. Cursor has rebuilt its systems and advised users to check for compromise.
  5. The attack is part of a wider trend of targeting AI developer platforms.

Sources

  1. BleepingComputer

    Reference

The day's cyber news in 5 minutes.

Cyberend Daily is one short email with the stories that matter, written so anyone can follow.

Includes our free beginner roadmap · No spam · Unsubscribe anytime