Breach
Cursor confirms breach that let hackers push malicious Claude Code extension
Cursor confirmed in an incident report published Friday that attackers stole an access key from Fireworks AI and used it to reach internal systems and steal customer data, including credentials.
Written by a model from the sources below. Verify against them.
- Vendor
- Cursor
- Product
- Cursor
- Status
- Confirmed
- Disclosed
- Data taken
- About 5K records: Source code, Tokens and keys, Email addresses
- BleepingComputer: “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 inferen…”
- Third-party vendor, Stolen credentials
- BleepingComputer: “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 inferen…”
- Fireworks AI
- BleepingComputer: “The incident comes amid a wave of attacks targeting AI developer platforms that cybersecurity company Wiz has dubbed "Black October" .”
- Black October
Summary
Cursor said malicious copies of Anthropic's Claude Code extension, versions 2.1.312 and 2.1.313, were served from its VS Code extension marketplace on October 29 for about 11 hours, and the article says up to 5,000 developers were exposed.
What to do
Rotate API keys and tokens you gave it, and review the OAuth apps and integrations connected to it
For tokens and keysExpect phishing that uses this breach; verify messages that mention it
For email addressesCheck whether your secrets or data were in its code
For source codeRemove Claude Code extension versions 2.1.312 and 2.1.313 and reinstall the extension from a verified source
Check developer machines for unexpected SSH keys and persistence mechanisms
Rotate any API credentials and secrets that may have appeared in AI coding-agent traces or support conversations
Read Cursor's security blog post for the scope of the breach and any required steps
Rotate credentials and API keys that were used with Cursor IDE
Timeline
- Made public by BleepingComputer
- First seen by cve.now
- Confirmed by Cursor
BleepingComputer: “Cursor confirmed it suffered a security breach that let hackers distribute a malicious Claude Code extension through its marketplace after stealing credentials…”
1 day - Record last updated
What was taken
About 5K records, perBleepingComputer
| Taken | Per the source |
|---|---|
| Source code, Tokens and keys | Catalin Cimpanu on Bluesky: “From these tools they accessed select archived non-production Cursor source code repositories (some containing business contact information), limited customer…” |
| Email addresses | BleepingComputer: “Limited customer support information and email conversations.” |
Evidence 10
- Social
- Social
- SocialBluesky
“Attackers accessed limited data via a leaked key from Fireworks, then placed a malicious VS Code extension for 11 hours.”
- PressBleepingComputer
“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.”
+1 more: BleepingComputer on Bluesky - SocialBluesky
“Cursor says the breach spread through stored data and internal credentials, potentially exposing up to 5,000 users to malicious Claude Code extensions.”
- PressCatalin Cimpanu on Bluesky
“Looks like the Cursor IDE breach was also caused by Fireworks AI... prolly this investigation led to Fireworks finding their own breach”
- SocialBluesky
“Cursor says attackers used a stolen Fireworks AI key to access internal data and credentials, then distributed malicious Claude Code extensions for about 11 hours.”
- SocialMastodon
“Cursor says an attacker used a key exposed by vendor Fireworks to copy coding traces, harvest customer credentials and reach internal tools before planting a tampered Claude Code extension on October 29.”
- SocialBluesky
“A few coding agent traces, shared for training, still held unredacted credentials, and those got the attacker into its SaaS tools.”
- SocialMastodon
“cursor-confirms-security-breach-led-to-malicious-claude-code-extension”
Corrections
- Changed “Tokens and keys were exposed” to “Tokens and keys of fewer than 0.01% of customers were exposed”.
- Changed “Tokens and keys of fewer than 0.01% of customers were exposed” to “Tokens and keys were exposed”.
Something wrong on this page? Report an error. Every fix is listed here.
Questions about the Fireworks AI breach affecting Cursor
Was Cursor customer data exposed?
Yes. The breach was at Fireworks AI, a vendor of Cursor. Cursor says so: “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.” (BleepingComputer).
What data was exposed?
About 5K records. Source code, Tokens and keys, Email addresses.
What should Cursor customers do?
Rotate API keys and tokens you gave it, and review the OAuth apps and integrations connected to it. Expect phishing that uses this breach; verify messages that mention it. Check whether your secrets or data were in its code. Remove Claude Code extension versions 2.1.312 and 2.1.313 and reinstall the extension from a verified source. Check developer machines for unexpected SSH keys and persistence mechanisms. Rotate any API credentials and secrets that may have appeared in AI coding-agent traces or support conversations. Read Cursor's security blog post for the scope of the breach and any required steps. Rotate credentials and API keys that were used with Cursor IDE.