Fireworks AI Says Hackers Accessed Some Customers' Datasets and Incomplete Model Weights
An update to its October breach notice adds account and billing data, plus a second route in through a third-party cloud provider; Fireworks says payment card numbers were not affected.
Published
Last updated / Fact-checked on
Illustrative photo of yellow Ethernet cables plugged into a network switch in a server rack; it is not Fireworks AI, its platform, or any customer named in this article.
Fireworks AI said on Monday, November 9, in an update to its October 9 security bulletin, that its investigation has found impacts to customers’ account and billing information and, for a limited set of accounts, to secrets, datasets and incomplete models. Fireworks says payment card numbers are not stored by it and were not impacted.
The update also describes a second route in: a third-party cloud provider that hosted Fireworks R&D machines had a security incident on October 23, and on October 24 an unauthorized party used it to reach Fireworks development machines and run a workload that “led to additional customer data impact,” Fireworks said. Fireworks did not name the provider.
At a glance
| Item | Detail | Source/status |
|---|---|---|
| Newly found impact | Account and billing information; for a limited set of accounts, secrets, datasets and incomplete models | Fireworks, Nov. 9 update |
| First path | Oct. 6: misuse of internal credentials; eval-job secrets in a limited number of accounts | Fireworks, Oct. 9/10 bulletin |
| Second path | Oct. 23 cloud-provider incident, used Oct. 24 to reach development machines | Fireworks, Nov. 9 update |
| Payment cards | Not stored; not impacted | Fireworks |
| Status | Investigation ongoing; findings “to date” | Fireworks |
| Customers | About 10,000 enterprise customers; names do not mean impact | Forbes |
What Fireworks says was accessed
Fireworks says the affected internal application provided access to resource metadata, such as identifiers, display names, descriptions and creation times, and to billing information, including billing and shipping addresses, invoice and payment history, account balances and credit usage. For certain customers, “a subset of incomplete model weights and datasets were accessed,” the company said.
Fireworks says it is notifying customers whose data was impacted and has not detected any impact to Inference, Training SDK and Nexus, which are zero data retention by default, adding: “These findings are based on our investigation to date. Our investigation is ongoing and may reveal additional information.” Fireworks has not named affected customers.
Two routes in, weeks apart
Fireworks’ original bulletin said that on October 6 it identified unauthorized use of internal Fireworks credentials. An unauthorized third party accessed environment variables used in evaluation jobs in a limited number of customer accounts, and Fireworks believed API keys and tokens stored as secrets were captured. At the time, it said it had not detected unauthorized access to customer models, datasets or training data.
The November 9 update adds the October 23–24 cloud-provider route, which Fireworks says it detected the same day. Fireworks says that “as of October 25,” it had not detected further evidence of successful unauthorized access by the threat actor on the platform customers use. The update does not say which data categories came through which path.
Fireworks says “the evidence indicates links to” a financially motivated cybercriminal group aimed at selling tokens from unauthorized access to large language models to illicit routing gateways, known as “LLMJacking.” It did not name the group. A November 2 Google Threat Intelligence Group/Mandiant blog describes LLMJacking generally, including fraudulent charges above $100,000 in a single day in some other organizations’ incidents; the blog does not mention Fireworks, and those figures are not Fireworks’ losses.
Bloomberg, in an article DailyMech could read only in part because of a paywall, called this Fireworks’ “second known hack in a matter of weeks” and quoted a spokesperson saying the company is “actively working with our customers to manage the incident, including Notion, Cursor and others.”
Forbes reported that two customers told it they had “substantially moved their business off Fireworks’ systems,” one describing “very bad hacks” of increasing frequency and criticizing Fireworks’ communications; two other customers told Forbes they were not aware of any hacks. Forbes says Fireworks counts roughly 10,000 enterprise customers, including Uber, Shopify, Upwork, DoorDash, Cursor and Cognition. Being named as a customer, by Forbes or Bloomberg, does not mean a company’s data was affected.
What Fireworks has changed
Fireworks says it notified law enforcement and that Palo Alto Networks Unit 42 is assisting with forensics. Its listed fixes include replacing credentials and service accounts, removing unnecessary elevated privileges, isolating affected systems and adding monitoring.
Self-serve dataset uploads and downloads for Batch Inference and dataset-based training are disabled while it investigates, Fireworks says. The Training SDK remains available with zero data retention, and Fireworks Lab-based training runs in an isolated environment, the company says.
What to watch
Further Fireworks updates as the investigation continues; whether the cloud provider is identified; when dataset uploads and downloads return; and whether any affected customer confirms impact.
How we reported this
Incident details come from Fireworks’ two bulletins, both fully read. Customer figures and customer accounts are Forbes’ reporting, fully read. Bloomberg was read only in part.
Published and fact-checked November 9, 2026.
Sources
- Fireworks AI, “Update on October 2026 security incident,” last updated November 9, 2026, https://fireworks.ai/security/bulletins/october-2026-security-incident-update, fully read
- Fireworks AI, “Fireworks October 2026 security incident,” published October 9, 2026, 10 PM PDT, last updated October 10, 2026, 11:16 AM PDT, https://fireworks.ai/security/bulletins/october-2026-security-incident, fully read
- Google Threat Intelligence Group / Mandiant, “The Rise of LLMJacking: How Threat Actors Monetize Compromised Cloud and Model Resources,” November 2, 2026, https://cloud.google.com/blog/topics/threat-intelligence/llmjacking-monetizing-compromised-cloud-models, fully read; does not mention Fireworks
- Anna Tong and Rashi Shrivastava, “Fireworks Cyberattack Exposes Customer Data And AI Models As Fallout Widens,” Forbes, November 9, 2026, 1:53 PM EST (10:53 AM PST), https://www.forbes.com/sites/annatong/2026/11/09/fireworks-cyberattack-exposes-customer-data-and-ai-models-as-fallout-widens/, fully read
- “Nvidia-Backed Fireworks AI Discloses Its Second Hack in Weeks,” Bloomberg (via Bloomberg Government), November 9, 2026, 8:46 PM UTC (12:46 PM PST), https://news.bgov.com/us-law-week/nvidia-backed-fireworks-ai-discloses-its-second-hack-in-weeks, partly read (first three paragraphs; rest paywalled)
DailyMech used AI tools to help research and draft this article; it was fact-checked by our editors.