Fireworks AI breach exposes customer billing data and model weights

Fireworks AI disclosed on Monday that its investigation into a security incident first reported last month has uncovered further malicious activity. The update indicates that the breach affected account and billing information, secrets, datasets and incomplete models for a limited set of accounts. In the earliest notice, Fireworks said unauthorised use of internal credentials allowed a third party to access environment variables used in evaluation jobs in a limited number of customer accounts.

That activity was detected on 6 October 2026 at 09:00 PDT. At that time the company said the only evidence showed environment variables, including API keys and tokens stored as secrets, were captured, with no signs that customer models, datasets, training data, inference traffic or personal information were accessed.

The latest bulletin broadens the scope of impact. Fireworks said attackers gained entry to its development environment via an unnamed third-party cloud provider hosting Fireworks R&D machines on 23 October 2026. The firm added that by 24 October the intruder used this access to run an unauthorized workload that led to additional customer data exposure, and by 25 October there was no evidence of further successful unauthorized access to the platform.

The exposed data types include resource metadata, billing information (billing and shipping addresses, invoice and payment history, account balance, credit usage) and a subset of incomplete model weights and datasets for certain customers. Fireworks emphasised that payment card numbers are not stored by Fireworks and were not impacted.

It said it has taken security-hardening steps, including credential revocation and token rotation, isolating affected systems, tightening access controls, and improving logging and monitoring. The company also linked the activity to a financially motivated cybercriminal group involved in selling tokens from unauthorised access to large language models, and noted the threat is associated with LLmjacking.