CURSOR.
DATA BREACH
Cursor (technology, the United States) was hit by a data breach first reported on 14 November 2026. No group has been publicly named as behind it in these reports. 7 news reports in the last 14 days, 2 in the last 24 hours; the latest, from Le Monde Informatique: “Cursor: a key stolen from Fireworks AI behind a cyberattack”. See every attack on the live map.
What happened in the Cursor data breach
Cursor, the AI code editor company in the United States, disclosed a security breach that ended with a tampered copy of Anthropic's Claude Code extension on its VS Code extension marketplace. The malicious builds, versions 2.1.312 and 2.1.313, were live for about 11 hours on 29 October 2026, according to Cursor's security notice and reporting by BleepingComputer. Cursor says at most 5,000 users who requested those cached versions may have received them.
The breach traces back to an access key that Cursor had issued to Fireworks AI, an inference provider. Cursor said in an incident update published on 13 November 2026 that the key was exposed on Fireworks' side. The attacker used it to copy a read-only storage bucket holding coding-agent traces, mostly from 2025 and from sessions outside Privacy Mode. Some of those traces contained credentials that had not been removed, and the attacker used them to reach selected software tools that Cursor uses. Cursor's SaaS tools exposed archived non-production repositories, limited support data, usage metrics, some conversation summaries and credentials that users had sent to support, according to Network Security Magazine.
Cursor said the malicious build tried to install a persistent SSH backdoor on developers' computers. Security researchers at The Hacker News and Dark Reading described the code as a loader that could also send host details. Cursor says it found no evidence that Privacy Mode users, customer code repositories, its production environment, the Cursor app and official updates, or other marketplace extensions were affected. Cursor detected the last malicious activity on 1 November, and it says that activity was blocked.
Nobody has been named as the attacker. Fireworks AI's own security bulletin links its incident to a financially motivated group and to criminals selling unauthorised model access, according to a DEV Community analysis, but that attribution concerns the Fireworks incident rather than a named individual or group behind the Cursor intrusion.
Cursor removed the tampered extension, took the marketplace offline, worked with the hosting provider to take down the attacker's server, rotated keys, deactivated the Fireworks key and moved inference traffic off Fireworks. Anthropic separately published its legitimate Claude Code extension as version 2.1.314, according to SecurityWeek. Cursor says it is still monitoring its systems.
- What was taken: Cursor says an attacker copied data from a read-only storage bucket shared with Fireworks AI, including historical coding-agent traces. Using credentials found in those traces, the attacker reached some of Cursor's SaaS tools and gained access to archived non-production code, limited support data, usage metrics and some conversation summaries. Cursor also confirmed that customer API credentials and internal systems were exposed, according to BigGo Finance's account of the update. On 29 October, the attacker placed modified copies of Anthropic's Claude Code extension in the cache that serves the Cursor marketplace.
- Who is affected: Cursor says at most 5,000 users who requested specific cached versions of the extension on 29 October potentially received the malicious build. Cursor said fewer than 0.01% of customers had API credentials exposed, as reported by BigGo Finance.
- Who is behind it: Nobody has been named by Cursor or the reporting sources. Fireworks AI's bulletin links its own incident to financially motivated criminals selling unauthorised model access, a practice Google Threat Intelligence Group calls LLMJacking. That link is Fireworks' assessment and is not a confirmed attribution of the Cursor intrusion.
- The response: Cursor removed the tampered Claude Code build, took its extension marketplace offline and rebuilt it, and had the cloud provider hosting the attacker's server take it down. It rotated keys and internal credentials, revoked and deactivated the Fireworks access key, cut off Fireworks' access to shared datasets, and moved all inference traffic off Fireworks. It also expanded secret scanning and network restrictions and notified affected users. Cursor published a security notice on 29 October and an incident update on 13 November 2026. Fireworks AI said it is working with Palo Alto Networks Unit 42 and that its investigation is ongoing.
Was my data taken in the Cursor data breach?
For this data breach: Cursor says uninstalling the extension alone may not be enough. Check your computer for unexpected SSH keys and persistence, and update to the legitimate Claude Code version 2.1.314 or later. If you sent credentials to Cursor support or used Cursor without Privacy Mode, rotate any API keys or secrets you shared.
Only Cursor can confirm whose data was involved. If you are a customer, these five steps cover you either way:
- Look for an email or letter from Cursor. Firms whose data is taken usually have to tell the people affected, and it says what was taken.
- If you have an account with Cursor, change its password, and the password anywhere else you used the same one.
- Turn on two-step sign-in where you can, so a stolen password is not enough on its own.
- Expect scam emails, texts and calls that mention Cursor or this breach. Do not click links in them; go to the real website yourself.
- Check whether your email address has turned up in known breaches at haveibeenpwned.com. In the UK, report money lost to fraud to Action Fraud.
Cursor data breach: the details
- Type: Other attack
- Sector: Technology
- Where: San Francisco, United States
- Reports, last 14 days: 7
- Latest report: 07:39 UTC, 16 November 2026
See the numbers
| Day | Reports |
|---|---|
| 3 Nov | 0 |
| 4 Nov | 0 |
| 5 Nov | 0 |
| 6 Nov | 0 |
| 7 Nov | 0 |
| 8 Nov | 0 |
| 9 Nov | 0 |
| 10 Nov | 0 |
| 11 Nov | 0 |
| 12 Nov | 0 |
| 13 Nov | 0 |
| 14 Nov | 3 |
| 15 Nov | 2 |
| 16 Nov (so far) | 2 |
Latest Cursor news
- Cursor: a key stolen from Fireworks AI behind a cyberattackLe Monde Informatique · Other attack · San Francisco, United States · 16 Nov 2026, 07:39 UTC
- Cursor discloses internal compromise after credentials leaked from AI coding history; explains how malicious 'Claude Code' extension was distributed合同会社ロケットボーイズ · Data breach · San Francisco, United States · 16 Nov 2026, 04:30 UTC
- Stolen Fireworks AI key led to Cursor data breach and trojanized Claude Code extensionThe Hacker News · Data breach · San Francisco, United States · 15 Nov 2026, 08:22 UTC
- Stolen Fireworks AI key led to Cursor data breach and trojanized Claude Code extensionThe Hacker News · Data breach · San Francisco, United States · 15 Nov 2026, 08:22 UTC
- Cursor confirms security breach led to malicious Claude Code extensionBleepingComputer · Other attack · Place unknown · 14 Nov 2026, 17:27 UTC
- Cursor: cyberattack originating from a contractor affects usersFuites Infos · Data breach · Place unknown · 14 Nov 2026, 13:23 UTC
- Cursor hit by cyberattack: hackers infiltrate users' computersCyberattaque.org · Other attack · San Francisco, United States · 14 Nov 2026, 11:21 UTC
Cursor data breach: timeline
- 2026-10-29: Malicious copies of Anthropic's Claude Code extension, versions 2.1.312 and 2.1.313, were placed in the Cursor VS Code extension marketplace cache. Cursor published a security notice the same day.
- 2026-11-01: Cursor says it detected the last malicious activity by the attacker, which was blocked.
- 2026-11-10: Fireworks AI disclosed the breach in which the Cursor-issued access key was exposed, according to a DEV Community analysis.
- 2026-11-13: Cursor published an incident update linking the Fireworks key exposure to the malicious Claude Code extension and the data breach.
- 2026-11-14: BleepingComputer and other outlets reported Cursor's confirmation of the breach and the malicious extension.
- 2026-11-15: The Hacker News and Network Security Magazine reported Cursor's disclosure that a stolen Fireworks AI key led to the data breach and the trojanised extension.
Sources for the story
- Cursor confirms security breach led to malicious Claude Code extension
- Malicious Claude Code Extension Served in Cursor Supply Chain Attack - SecurityWeek
- Devs Hit by Trojanized Claude Code Extension for Cursor
- Official Claude Code Extension on Cursor Marketplace Trojanized with Persistent SSH Backdoor
- Cursor Reveals How a Leaked Fireworks AI Key Led to a Backdoored Claude Code Extension
- Cursor Traces Breach to Access Key Exposed by Fireworks AI
- Cursor Says Stolen Fireworks AI Key Led to Trojanized Claude Code Extension Incident
- Cursor says hackers used a key exposed at Fireworks AI to plant a malicious Claude Code extension in its marketplace
- Cursor Confirms a Stolen Fireworks Key Let Attackers Ship a Backdoored Claude Code Extension to 5,000 Developers - DEV Community
- The Secrets Were in the Training Data: What the Cursor Breach Actually Teaches - DEV Community
Cursor data breach: common questions
What happened to Cursor?
Cursor (technology, the United States) was hit by a data breach first reported on 14 November 2026. No group has been publicly named as behind it in these reports.
Was my Cursor account affected by the Claude Code extension breach?
Cursor says at most 5,000 users who requested the affected cached versions on 29 October may have received the malicious build.
Did the attack reach Cursor's customer code repositories?
Cursor says it has found no evidence that customer code repositories, Privacy Mode users, its production environment, the Cursor app or its official updates were affected. The attacker did reach archived non-production repositories through its SaaS tools, according to Network Security Magazine.
Is this the same as the Fireworks AI breach?
DEV Community describes the Cursor incident as a follow-up to the Fireworks AI breach rather than the same event. Fireworks exposed the key, and the attacker then used it to copy data and plant the malicious extension at Cursor.
Who carried out the attack?
No attacker has been named by Cursor. Fireworks AI links its own incident to a financially motivated group, but that is its assessment and not a confirmed attribution of the Cursor intrusion.
Is the malicious Claude Code extension still available?
Cursor removed the modified extension from the store after identifying it and took the marketplace offline to investigate. Anthropic published its legitimate Claude Code extension as version 2.1.314, according to SecurityWeek.
Who is behind the Cursor data breach?
No group has been publicly named in the reports collected here. Attribution can take weeks and is often disputed.
Was my data affected?
Only Cursor can say whose data was involved; its own statement is usually linked from the latest reports below. If you are a customer, watch for emails or texts that mention the breach, change your password there and anywhere you reuse it, and turn on two-step sign-in.