Fireworks AI breach exposed customer data and incomplete model assets
Fireworks AI said attackers accessed environment variables in evaluation jobs, billing information, and a subset of some customers’ incomplete model weights and datasets. Notion separately said the incident affected fewer than 1% of its workspaces and led it to disable an AI feature while investigating.

Key takeaways · 4
- 01
Fireworks said billing and shipping addresses, invoice and payment history, account balances, and credit usage were exposed; it said stored payment card numbers were not affected because it did not store them.
- 02
The company said attackers accessed a subset of certain customers’ incomplete model weights and datasets.
- 03
Fireworks disabled self-service dataset uploads and downloads for Batch Inference and dataset-based training during its investigation.
- 04
Notion said it revoked exposed Notion API keys on November 11 and reported exposed third-party API keys to their vendors on November 12.
Two stages in Fireworks’ account
Fireworks said it identified unauthorized use of internal credentials on October 6, with a third party accessing environment variables in evaluation jobs across a limited number of customer accounts.[1] The company also described an incident at a third-party cloud provider hosting its research and development machines on October 23; it said an unauthorized party used the provider to reach Fireworks development machines the next day.[2] Fireworks said an unauthorized workload caused additional customer data impact and was detected the same day.[2] The company did not attribute every exposed item to a particular stage in these statements.
What Fireworks says was exposed
Fireworks said billing and shipping addresses, invoice and payment history, account balances, and credit usage were exposed.[2] It said payment card numbers were not stored by the company and were not affected.[2] The company also said attackers accessed a subset of incomplete model weights and datasets belonging to certain customers.[2] Fireworks said it had not detected evidence of unauthorized access to customer models, datasets, training data, personal information, or inference traffic within its security perimeter or control.[1] It said no impact was found to Inference, Training SDK, or Nexus, which are zero-data-retention by default.[2]
Notion’s response and user data
Notion said fewer than 1% of its workspaces were affected, and that the dataset included chat transcripts from its AI Insights feature.[3] It said most impacted workspaces could not be identified from those transcripts, which were not labeled with the writer or workspace and did not contain tool-call results.[3] Notion said it disconnected its systems from Fireworks by the morning of November 3, disabled Insights during its investigation, and stopped serving open-weight models through Fireworks.[3] It later revoked exposed Notion API keys found in affected user messages and reported exposed third-party API keys to the issuing vendors.[3]
Containment and broader risk context
Fireworks said it revoked and replaced credentials and service accounts, removed unnecessary elevated privileges, and tightened controls on credential creation.[2] It also said it restricted unauthorized network pathways, isolated affected systems, and disabled self-service dataset uploads and downloads for Batch Inference and dataset-based training while investigating.[2] As of October 25, the company said it had detected no further evidence of successful unauthorized access to its customer-facing platform.[2] TechCrunch described this as Fireworks’ second breach in as many months.[4] Separately, Google Threat Intelligence Group and Mandiant said LLMJacking incidents in 2026 affected organizations across several regions and industries, with some organizations facing charges exceeding $100,000 in a day.[5]
The disclosures show why enterprise AI reviews should cover both the model provider and the cloud or service dependencies used to build and operate AI systems. Teams should also distinguish exposed metadata or incomplete assets from the customer data and services a provider says were not affected.
Why it matters
AI training for your team
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
How this developed
15 November 2026
Fireworks AI breach exposed customer data and incomplete model assets
Sources
- Fireworks October 2026 security incident | Fireworks AIfireworks.ai
- Update on October 2026 security incident | Fireworks AIfireworks.ai
- Fireworks AI incidentnotion.com
- Cyberattack on Fireworks AI allowed hackers to access Notion customers' chat transcripts | TechCrunchtechcrunch.com
- The Rise of LLMJacking: How Threat Actors Monetize Compromised Cloud and Model Resources | Google Cloud Blogcloud.google.com