Fireworks Breach Hit Datasets, Partial Models

Fireworks AI said on Monday, Nov. 9, that attackers who broke into its systems in October reached customer billing records and, for a limited set of accounts, datasets and partly trained model weights. That is wider than its first disclosure on Oct. 9, which covered only stored secrets.

Key takeaways

• Fireworks now says account and billing data were exposed, plus secrets, datasets and incomplete model weights for some customers. Payment card numbers were not stored or affected.
• A second intrusion on Oct. 24 came through a third-party cloud provider that hosted Fireworks research machines.
• The company ties the attacks to a financially motivated group that resells stolen AI access, a practice known as LLMjacking.

What the Fireworks breach update says

In its updated security bulletin, Fireworks said it has since found impacts to account and billing information beyond the secrets stored for evaluation jobs.

Exposed data includes resource metadata and billing details such as billing and shipping addresses, invoice and payment history, account balances and credit usage. For certain customers, a subset of incomplete model weights and datasets was also accessed.

Fireworks said it has not detected any impact to its Inference, Training SDK and Nexus products, which keep no customer data by default. It cautioned that its investigation is ongoing and "may reveal additional information."

A second entry point on Oct. 24

The bulletin describes a separate incident. A third-party cloud provider that hosted Fireworks research and development machines was breached on Oct. 23. The next day, the intruder used that access to reach Fireworks development machines and run an unauthorized workload, which caused further customer data impact.

Fireworks said it detected the activity the same day. As of Oct. 25, it had found no further evidence of successful unauthorized access on the platform customers use.

How Fireworks is responding

The company said it notified law enforcement and brought in Palo Alto Networks' Unit 42 for forensic work. It replaced credentials, cut unneeded privileges and isolated affected systems.

Fireworks has also turned off self-serve dataset uploads and downloads used by Batch Inference and dataset-based training while the investigation continues. Its Training SDK remains available.

Customer fallout and the LLMjacking context

Forbes reported that Fireworks counts roughly 10,000 enterprise customers, including Uber, Shopify and DoorDash, and is valued at $17.5 billion. Two customers told the outlet they had substantially moved their business off Fireworks, with one describing "very bad hacks" and poor communication. Two others said they were unaware of any hacks.

Fireworks was the first of at least nine AI developer platforms that cloud security firm Wiz counted disclosing incidents in October. Wiz's running log has since added six more disclosures through Nov. 9, and it says the attacks share a hunt for stored credentials that grant AI access.

The financial stakes are large. In a report this month, Google Threat Intelligence Group and Mandiant said LLMjacking victims saw fraudulent charges above $100,000 in a single day, with cumulative bills at some organizations reaching $1 million before access was cut off. One actor ran more than 500,000 model API calls in under eight hours.

What's next

Fireworks said it is notifying customers whose data was affected and will post further updates. Teams that store provider keys on hosted AI platforms should rotate them and set spending caps, as Google's researchers advise.

FAQ

Was my payment card exposed in the Fireworks breach?

Fireworks says it does not store payment card numbers and that they were not affected. Billing addresses, invoices and payment history were in scope.

What is LLMjacking?

It is the theft of access to AI models, often via stolen API keys or cloud credentials. Attackers resell the access while the victim pays the bill.

What should Fireworks customers do now?

Follow any notice from Fireworks, rotate keys and third-party credentials stored on the platform, and review usage and billing logs for unfamiliar activity.

References

• Update on October 2026 security incident — Fireworks AI (Nov. 9, 2026)
• Fireworks October 2026 security incident — Fireworks AI (Oct. 9, 2026)
• Fireworks Cyberattack Exposes Customer Data And AI Models As Fallout Widens — Forbes (Nov. 9, 2026)
• "Black October" for AI Developer Platforms — Wiz (Oct. 30, 2026, updated)
• The Rise of LLMJacking: How Threat Actors Monetize Compromised Cloud and Model Resources — Google Cloud Blog (Nov. 2026)