Fireworks Expands Disclosure of October Security Incident

Cybersecurity team in a data center working to prevent unauthorized network access

Fireworks AI expanded its disclosure of an October security incident in a bulletin updated November 9, 2026. The company says its investigation found impacts to account and billing information beyond the stored secrets described in its initial October 9 notice. For a limited set of accounts, it also identified impacts involving secrets, datasets and incomplete models.

Additional activity and scope

According to Fireworks, a third-party cloud provider hosting its research and development machines experienced an incident on October 23. The company says an unauthorized party reached development machines the following day and ran a workload that led to additional customer-data impact. Fireworks says it detected that activity on October 24 and, as of October 25, had not detected further successful unauthorized access on its customer-facing platform.

What the company says was affected

The bulletin lists resource metadata, billing information and, for certain customers, a subset of incomplete model weights and datasets. Fireworks says it does not store payment-card numbers and those numbers were not impacted. It also says it has not detected impact to Inference, Training SDK or Nexus, which it describes as zero-data-retention services by default. These are company findings from an ongoing investigation, not an independent assurance that the scope cannot change.

Response and customer action

Fireworks reports working with Palo Alto Networks Unit 42, replacing credentials, restricting network pathways and strengthening storage-access and logging controls. It has disabled self-service dataset uploads and downloads used by Batch Inference and dataset-based training during the investigation. Customers should use their individual incident notifications to assess exposure; the general bulletin does not establish that every account or workload was affected.

Why it matters

The incident illustrates why AI-service risk assessments need to distinguish production inference from development infrastructure, stored training material and account-management data. Fireworks links the activity to financially motivated token resale described as LLMJacking. Attribution and impact statements here remain explicitly attributed to the company.

Source and verification

Fireworks AI security bulletin, updated November 9, 2026. Reviewed on November 15, 2026. Source date is shown separately from publication time.

More technology news

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *