The Pulse
Fireworks AI Says Intruder Captured Secrets in Evaluation Jobs
Fireworks AI says an unauthorized third party accessed environment variables in evaluation jobs across a limited number of customer accounts. The company has revoked identified keys and tokens, while asking affected customers to rotate thir
AI.info Team · Written with AI ·
“We believe that environment variables, including API keys and tokens stored as secrets, were captured in certain affected accounts.”
Fireworks AI security bulletin
Secrets exposed through evaluation jobs
Fireworks AI says an unauthorized third party accessed environment variables used in evaluation jobs in a limited number of customer accounts. The company identified the activity on October 6, 2026, and began containment that day. Its bulletin, last updated October 10, lists its status as active monitoring and says the company has directly notified every customer it has identified as affected.
The exposed variables included secrets such as API keys and tokens, according to the company. Fireworks has not disclosed how many accounts were affected or how the unauthorized party gained access to its internal credentials. The bulletin also does not say whether the captured customer credentials were used outside Fireworks.
Fireworks reports no detected access to models or inference traffic
Fireworks says it has not detected evidence that customer models, datasets, training data, personal information or inference traffic were accessed within its security perimeter or control. That finding is qualified: the company is still reviewing calls made with the exposed credentials to verify the scope, and says its review of other potential access continues.
The company says its platform does not store prompts or outputs by default under its zero-data-retention policy. That statement concerns inference traffic on Fireworks; it does not establish whether an exposed third-party credential could reach data held by the provider that issued it. For that reason, Fireworks tells affected customers to ask those providers to review access logs and any data the credentials could reach.
Credential revocations continued through October 10
Fireworks’ timeline says it began notifying affected customers on October 7, rotated employee API keys on October 9 and notified identified customers by that evening. On October 10, the company revoked additional Fireworks user API keys and rotated service-account tokens found in a completed credential review. It says further review and rotation are continuing.
The company says it has revoked identified internal credentials used in the incident, affected Fireworks API keys and keys it identified as created by the unauthorized party. It has also added access restrictions and is reviewing how evaluation jobs receive secrets. Fireworks says it will share further remediation measures as they are implemented and verified.
Customers must rotate third-party keys themselves
Fireworks says affected customers should replace revoked Fireworks API keys if they still need them, then update applications that used those keys. Third-party credentials require a separate step: customers must revoke or rotate them with the issuing providers and update the applications and Fireworks secrets that rely on them. Removing a secret from Fireworks alone does not invalidate the credential.
The company also recommends checking for unfamiliar API keys, users, evaluators, training jobs and datasets, and reviewing provider logs for activity involving exposed third-party credentials. Customers who have not received an incident notice do not need to take incident-specific action at this time, Fireworks says. The company directs customers who see suspicious activity or need help confirming their status to reply to their notification email or contact its security team.