Fireworks AI says an attacker came in through a cloud provider and reached some customers' datasets and incomplete model weights, its second known hack in weeks

Fireworks AI, which runs open models for developers, widened its account of an October security incident in a bulletin last updated 9 November. “A third-party cloud provider that hosted Fireworks R&D machines experienced a security incident on October 23,” it says. On 24 October the intruder used that access to reach Fireworks development machines.

Fireworks says the attacker took resource metadata and billing information, and that “for certain customers, a subset of incomplete model weights and datasets were accessed.” It says “Payment card numbers are not stored by Fireworks and were not impacted.” It notified law enforcement and is working with Palo Alto Networks' Unit 42. It links the attack to a financially motivated group that sells tokens from stolen model access to illicit routing gateways, known as “LLMJacking”. It names no customers and gives no count.

Bloomberg, as published by Insurance Journal, called it Fireworks's “second known hack in a matter of weeks” and said customers were told on Monday; a person familiar told Bloomberg tens of thousands of files were taken. The Record reports account and billing information, secrets and datasets were stolen. Fireworks's earlier bulletin, for an incident in early October, said customer models and datasets had not been accessed.

• Confirmed An attacker who breached a cloud provider on 23 October reached Fireworks development machines on 24 October. Fireworks AI
• Claimed For certain customers, a subset of incomplete model weights and datasets were accessed; payment card numbers were not affected. Fireworks AI
• Reported It is Fireworks's second known hack in a matter of weeks, Bloomberg reported. Insurance Journal (Bloomberg)