Published

Fireworks updates security incident: billing data and some customers’ datasets and incomplete model weights were accessed

AI inference platform Fireworks updated its October security bulletin on 9 November, saying the breach went further than first disclosed: account and billing information was affected, and for a limited set of accounts secrets, datasets and incomplete model weights were accessed. It says payment card numbers are not stored and were not affected, that no impact has been found on Inference, Training SDK or Nexus, and that it has disabled self-serve dataset uploads and downloads.

Fireworks says that when it published its first bulletin on 9 October, only secrets it stored (for example for evaluation jobs) were known to have been accessed. Since then, investigating with Palo Alto Networks Unit 42, it found impacts to account and billing information, including billing and shipping addresses, invoice and payment history, account balances and credit usage, and for a limited set of accounts to secrets, datasets and incomplete model weights. It has notified law enforcement and is notifying affected customers.

According to the bulletin, the evidence points to a financially motivated crime group that sells tokens from unauthorized LLM access to illicit routing gateways, known as LLMJacking. A third-party cloud provider hosting Fireworks R&D machines had a security incident on 23 October, and on 24 October the attacker used it to reach Fireworks development machines and run an unauthorized workload, causing the additional customer data impact. Fireworks says it detected and responded the same day and has seen no further successful access to the customer platform since 25 October.

Fireworks says it does not store payment card numbers and they were not affected, and that no impact has been found on Inference, Training SDK or Nexus, which are zero data retention by default, though the investigation is ongoing. It has rotated credentials, isolated affected systems and tightened storage access, and has disabled self-serve dataset uploads and downloads for Batch Inference and dataset-based training during the investigation.

Sources

  1. Update on October 2026 security incident — Fireworks AI
Open this story in Shift →