// AI threat desk · 15 Nov 2026
Home/LEAK · Data Leaks
Data LeaksMediumAISECCVSS not assigned

Fireworks disclosed customer data access after October incidents

The Record reported that hackers gained access to internal Fireworks data, billing information, invoices, payment history, datasets and other AI-related data. Fireworks said a limited number of customer accounts were affected. Fireworks published its report Oct 10, 2026.

· 2 min read
AI-generated image of a silhouetted analyst reviewing blurred customer account and incident data on security monitors.
AI-generated image, not a photo of the event.
Key takeaways
  • The Record reported that hackers gained access to internal Fireworks data, billing information, invoices, payment history, datasets and other AI-related data.
  • The incidents affected a limited number of customer accounts, and Fireworks said all affected customers were notified.
  • Fireworks revoked identified credentials and keys and added access restrictions.

Fireworks said an unauthorized third party accessed customer data during incidents in October, including billing information, invoices, payment history, datasets and other AI-related data. The company said the incidents affected a limited number of customer accounts and that every affected customer was notified.

The access began with unauthorized use of internal Fireworks credentials identified on October 6. Fireworks said the third party later used the provider on October 24 to reach development machines and run an unauthorized workload that caused additional customer data impact.

The affected data included environment variables used in evaluation jobs. Fireworks said its initial investigation had not detected unauthorized access to customer models, training data, personal information or inference traffic.

On this page

Fireworks linked the access to credentials and a development workload

Fireworks identified unauthorized use of internal credentials on October 6 and began containment the same day. It said an unauthorized third party accessed environment variables used in evaluation jobs in a limited number of customer accounts.

Exposed data included secrets, datasets and incomplete model weights

Fireworks said the exposed information included account and billing details, addresses, payment records and some customer datasets. Separate reporting attributed to Fireworks said attackers reached stored secrets, datasets and incomplete model weights for a limited set of accounts.

Fireworks said it does not store payment card numbers, so card numbers were not affected. The company said its initial investigation had not found unauthorized access to customer models, training data, personal information or inference traffic.

The incidents fit a wider pattern of LLM access abuse

Mandiant said it responded to multiple incidents in 2026 in which threat actors hijacked enterprise access to large language models, a tactic known as LLMJacking.

Google Threat Intelligence Group said some LLMJacking incidents produced fraudulent charges exceeding $100,000 in a single day, with cumulative charges at some organizations reaching $1 million before access was revoked. Google Cloud published that report on Nov 2, 2026.

Fireworks said it revoked the internal credentials used in the incident, affected Fireworks API keys and keys created by the actor. It also added access restrictions and began containment on October 6.

Fireworks said it had directly notified every identified affected customer. Separate reporting said the company had seen no successful access to its customer platform since October 25.

What to do now

  1. Replace revoked Fireworks API keys if you still need them, then update applications using them.
  2. Rotate or revoke third-party credentials identified in your notice with the providers that issued them, then update the applications and Fireworks secrets that use them.
  3. Review data and activity associated with exposed third-party credentials.
  4. Check for unrecognized API keys, users, evaluators, training jobs and datasets.

FAQ

What happened at Fireworks?

Fireworks said an unauthorized third party used internal credentials and later reached development machines, exposing billing information, secrets, datasets and other AI-related data.

Who was affected by the Fireworks incident?

Fireworks said a limited number of customer accounts were affected and that every identified affected customer was notified.

Did attackers access Fireworks customer models?

Fireworks said its initial investigation had not detected unauthorized access to customer models, training data, personal information or inference traffic.

How should Fireworks customers respond?

Fireworks instructed affected customers to replace revoked API keys, rotate or revoke identified third-party credentials, review associated activity and check for unrecognized resources.

What is LLMJacking?

Mandiant described LLMJacking as a tactic in which threat actors hijack enterprise access to large language models.

Sources

  1. Fireworks October 2026 security incident, FireworksPrimary
  2. The Rise of LLMJacking: How Threat Actors Monetize Compromised Cloud and Model Resources | Google Cloud Blog, Google CloudPrimary
  3. AI company Fireworks says hackers stole data, billing info during October breaches, The Record
  4. Fireworks Cyberattack Exposes Customer Data And AI Models As Fallout Widens, Forbes
  5. Fireworks Says Hackers Accessed Customer Data and Unfinished AI Models, TokenPost
  6. Fireworks AI Hack Exposed Customer Datasets and Billing Data, Fireworks
How we checked this story
ClaimSourceStatus
Fireworks said an unauthorized third party accessed environment variables in a limited number of customer accounts.FireworksConfirmed
Fireworks said it had not detected unauthorized access to customer models, datasets, training data, personal information or inference traffic in its initial investigation.FireworksConfirmed
Fireworks said it directly notified every identified affected customer.FireworksConfirmed
Fireworks revoked internal credentials, affected API keys and actor-created keys, and added access restrictions.FireworksConfirmed
Fireworks said the incident affected a limited number of customer accounts and that all affected customers were notified.Fireworks, via The RecordAttributed
ShiftDelete reported that Fireworks said attackers reached billing records and, for limited accounts, secrets, datasets and incomplete model weights.FireworksAttributed
ShiftDelete reported that Fireworks said payment card numbers were not affected because the company does not store them.FireworksAttributed
ShiftDelete reported that Fireworks had seen no successful access to its customer platform since October 25.FireworksAttributed
Forbes reported that two customers had moved their business away from Fireworks, with one describing very bad hacks and poor communication.ForbesAttributed
Mandiant said it responded to multiple LLMJacking incidents in 2026 across cloud and LLM providers and several regions and industries.Mandiant, via Google CloudAttributed
Google Threat Intelligence Group said LLMJacking incidents produced fraudulent charges exceeding $100,000 in one day and reaching $1 million cumulatively at some organizations.Google Threat Intelligence Group, via Google CloudAttributed
fireworks.ai published its report on Oct 10, 2026.fireworks.aiConfirmed
Google Cloud published its report on Nov 2, 2026.Google CloudConfirmed

Could not verify

  • Whether attackers accessed customer models or training data is not established by the initial Fireworks investigation.
  • Whether the incidents involved successful access to Fireworks customer systems after October 25 is not established.
  • How many customer accounts and datasets were affected is not established.
  • Whether the two October incidents were conducted by the same cybercriminal group is not established.
  • Whether the attackers used AI agents is not established.
See all in Data Leaks →
Explore with AI
ChatGPT
Claude
Perplexity
Grok