Fireworks AI Breach Reaches Uber, Cursor and Notion, With 149,346 People Notified
Security / news
Fireworks AI Breach Reaches Uber, Cursor and Notion, With 149,346 People Notified
Uber's alerts never fired because the intruders used Fireworks' own credentials; Cursor traces its incident to a key it issued to the vendor.
Uber is notifying 149,346 people that their names, addresses, dates of birth and driver's licence numbers were taken in a breach of Fireworks AI, the inference provider it used, according to an Oregon attorney general filing cited by TechCrunch. Uber said it learned of the incident from the provider on October 9, and Fireworks did not answer TechCrunch's specific questions about the incident.
The same Fireworks intrusion has now reached at least three customers. Notion and Cursor have confirmed that their data was stolen too, TechCrunch reports, and Cursor's November 13 update traces its own incident to an access key it had issued to Fireworks.
What Fireworks says happened
Fireworks' security bulletin, updated November 9 and marked "Active monitoring", gives a two-stage account. The original bulletin on October 9 covered only Fireworks-stored secrets used in evaluation jobs. On October 23 a third-party cloud provider that hosted Fireworks R&D machines had its own incident, and on October 24 an unauthorized party used it to reach Fireworks development machines and run a workload that exposed more customer data. Fireworks detected that the same day. The bulletin does not name the cloud provider.
Data classes Fireworks lists: resource metadata (identifiers, names, descriptions, creation times); billing and shipping addresses, invoice history, account balance and credit usage; and, for a limited set of accounts, secrets, datasets and a subset of incomplete model weights. Fireworks says payment card numbers are not stored and were not affected, and that Inference, the Training SDK and Nexus, which run with zero data retention by default, were not impacted so far.
Fireworks attributes the activity to a financially motivated group that sells tokens from unauthorised model access, which Fireworks calls LLMjacking. It has notified law enforcement and hired Palo Alto Networks Unit 42.
Why Uber's alerts stayed silent
Uber told TechCrunch it had alerts for unauthorized access to its Fireworks environment. None fired, because the intruders used Fireworks' own internal credentials, so the activity looked like the vendor. A spokesperson described "a security incident impacting one of our third-party vendors, Fireworks.ai" and referred other questions to Fireworks.
Uber has filed with the California and Iowa attorneys general and has notified drivers in California, Illinois, Massachusetts, Rhode Island and South Carolina. It did not say how many were drivers rather than other people, and it is unclear whether the 149,346 figure will rise. This is Uber's second known breach in 2026, after the August claim by an extortion gang over Uber Freight.
| Customer | What was taken | Source |
|---|---|---|
| Uber | Name, address, date of birth, licence number for 149,346 people | Oregon AG filing, via TechCrunch |
| Cursor | Agent traces (mostly 2025), archived non-production repositories, support emails | Cursor, November 13 |
| Notion | Data confirmed stolen; the categories were not detailed in the Uber report | TechCrunch |
Cursor's exposure, in its own numbers
Cursor says the attacker started with the Fireworks-issued key, then used credentials found inside coding agent traces to reach its SaaS tools. Fewer than 0.05% of users not in Privacy Mode had traces exposed, and less than 0.01% of customers had API credentials exposed. Prompts and code in Privacy Mode were not affected.
On October 29 a malicious copy of Anthropic's Claude Code extension sat in Cursor's marketplace cache for about 11 hours. Cursor says at most 5,000 users who requested cached versions that day may have received it. Anyone who installed Anthropic.claude-code 2.1.312 or 2.1.313 from the Cursor marketplace should follow the steps in Cursor's October 29 notice, which warns that "Uninstalling the extension alone may not be enough." Cursor deactivated the Fireworks key and moved all inference off Fireworks.
For other Fireworks customers, the bulletin lists no response steps. It says affected customers will receive them in their notices, and directs questions to security-notice@fireworks.ai. Self-serve dataset uploads and downloads for batch inference and dataset-based training stay disabled during the investigation.
Two earlier Terminal reports on coding-agent tooling, Chrome DevTools MCP and Plannotator, cover other tools in that workflow that phone home or hold credentials of their own. Fireworks' next update is the one to watch for a final count of affected customers.
Sources
More in Security
- 01Java4Shell Scores 9.2 From Its Finder and 8.1 From OpenJDK, and Needs an AVX-512 CPUCalif says a 65-character Base64 string overwrites 62 bytes of Java heap and reaches code execution in Keycloak, Jenkins and Jira. A second security firm says the source shows only a crash.
- 02CISA Gave 9 of 16 November KEV Entries a Three-Day Deadline With Forensic Triage, and 7 Got 14 DaysThe catalog's own due-date and forensicTriage fields show two tiers under BOD 26-04. Sorting November's additions by tier shows which exploited bugs CISA treats as already-compromised-until-proven-otherwise.
- 03Project Zero Cuts Its Disclosure Deadline to 60 Days on Feb. 2, 2027, Then 30 in 2028Google's bug-hunting team is dropping the 30-day patch soak period and counting every deadline from the day it reports, citing AI-assisted exploitation.
- 04Google Finds exploit_agent Diffing Patches and Firing Exploits With No Operator in the LoopGTIG's November 16 tracker reports the first fully automated discovery-to-exploit pipeline it has seen, found on an exposed staging server, plus 22 zero-days in August.