Fireworks AI discloses second breach: attackers used a cloud provider hack to reach billing data and model weights

AI inference company Fireworks AI says attackers reached its development machines through a third-party cloud provider and accessed customer data, according to an update to the security bulletin it first published on 9 October about an earlier incident. The exposed information included resource metadata and billing details, while secrets, datasets and incomplete model weights were affected for a limited set of accounts. Payment card numbers were not affected because Fireworks AI does not store them.

The cloud provider experienced a security incident on October 23. On October 24, the attacker used that access to reach Fireworks AI development machines and run an unauthorized workload. Fireworks AI detected the activity that day and says it found no further successful unauthorized access after October 25. The company says the evidence indicates links to a financially motivated group involved in LLMJacking, which uses stolen access to language models and sells it through illicit routing services.

Fireworks AI revoked and replaced credentials, isolated affected systems, restricted network access and tightened access to data storage. It also disabled self-service dataset uploads and downloads for Batch Inference and dataset-based training. The company says it found no impact to Inference, Training SDK or Nexus, which use zero data retention by default, and continues to investigate with Palo Alto Networks Unit 42, BleepingComputer reports.