🔍 Read the full analysis: Fireworks AI Breach: Reviewing Datasets And Weights on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Fireworks AI’s November 9 update says an October security incident affected more information than its first disclosure indicated. The company reports access to account and billing information and, for a limited set of accounts, secrets, datasets and incomplete model weights; the number of affected customers and the fate of the data remain unknown.
Fireworks AI said on November 9 that an October security incident exposed more customer information than the company had initially identified, including account and billing details and, for a limited set of accounts, secrets, datasets and incomplete model weights. The update describes two access routes, including an unauthorized workload run through development machines hosted by a third-party cloud provider; Fireworks has not named that provider or disclosed how many customers were affected.
Fireworks, which provides infrastructure companies use to run and fine-tune open AI models, first disclosed suspicious activity on October 9. Its initial bulletin said internal credentials had been used to read environment variables from evaluation jobs for a limited number of customer accounts. At that point, the company said it had not detected unauthorized access to customer models, datasets, training data, inference traffic or personal information.
The November update broadened that account. Fireworks said the incident affected resource metadata and billing information, including billing and shipping addresses, invoice and payment history, account balances and credit usage. For a limited set of accounts, it also found that attackers accessed secrets, datasets and a subset of incomplete model weights. The company has not explained what “incomplete” means or specified how much data was involved.
Fireworks’ timeline describes a separate route through a third-party cloud provider that hosted its research and development machines. The provider experienced a security incident on October 23; Fireworks says an unauthorized party used that access on October 24 to reach development machines and run an unauthorized workload, causing additional customer data impact. The company says it detected the activity that day and has not detected further successful unauthorized access to the platform since October 25.
Fireworks AI breach expanded data review
A November 9 update broadened the scope of an October security incident. The number of affected customers and what happened to accessed data remain unknown.
Information Fireworks says was accessed
The reported scope includes account and billing details, with additional data for a limited set of accounts.
Metadata and billing
Resource metadata, addresses, invoice and payment history, account balances and credit usage.
Secrets, datasets and weights
Fireworks reports access to secrets, datasets and a subset of incomplete model weights.
Reported unaffected
Fireworks says it does not store payment card numbers and they were not affected.
How the scope widened
The update describes two access routes and a later third-party cloud provider incident.
What customers can and cannot conclude
The company has reported safeguards and product boundaries, while key details remain unresolved.
- No further successful platform access detected since October 25.
- No detected impact to Inference, Training SDK or Nexus, described as zero-data-retention by default.
- Dataset uploads and downloads for Batch Inference and dataset-based training are disabled during the investigation.
- How many customers were affected.
- Which cloud provider was involved and how initial credentials were obtained.
- Whether data was copied, retained, shared or used, and what “incomplete” weights contained.
Investigation and next steps
Fireworks says its review is ongoing with forensic assistance from Palo Alto Networks Unit 42, and that it has notified law enforcement.
- Customers need account-specific details and guidance on whether accessed secrets should be rotated.
- No restoration date for the disabled dataset functions or final report timetable has been provided.
- Notion separately said fewer than 1% of its workspaces were affected; that figure does not establish the wider Fireworks customer impact.
Risks for Customer Data and Models
The disclosure matters because Fireworks operates infrastructure used by other companies to process data and develop AI products. A dataset, secret or model artifact held by a service provider can contain material customers regard as sensitive or commercially valuable. The reported exposure could therefore affect customers beyond Fireworks itself, although the company’s update does not establish that any exposed information was used or published.
Model weights are learned parameters produced during model training. For a company that fine-tunes a model on its own data, those weights can form part of its product or intellectual property. Fireworks says only a subset of incomplete weights was accessed for certain customers; the practical significance depends on what was copied, how complete it was and whether the data can be used. Those details have not been provided.
Fireworks says it does not store payment card numbers and that they were not affected. It also says it has detected no impact to its Inference, Training SDK or Nexus products, which it describes as zero-data-retention by default. Those statements narrow the categories the company reports as affected, but its investigation remains open.
Products Worth Considering
LG UltraGear 34GX90SA-W 34-inch WQHD Curved OLED Gaming Monitor
LG UltraGear 27G640A-B 27-Inch QHD IPS Gaming Monitor, 300Hz, 1ms, G-SYNC Compatible, FreeSync Premium, DisplayHDR 400
Ledger Nano Gen5 Crypto Wallet, Black
As an affiliate, we earn on qualifying purchases.
From October Bulletin to Expanded Scope
Fireworks says it identified the first activity on October 6 and began containment. It started notifying affected customers on October 7, rotated employee API keys on October 9 and revoked additional user API keys and rotated service-account tokens on October 10. Its first public bulletin, dated October 9, described access to environment variables and said no evidence had been found of access to customer datasets, models or personal information.
The later cloud-provider incident changed the scope under review. Fireworks says the provider’s security incident occurred October 23 and that the unauthorized workload ran the following day. On November 9, the company said its investigation had identified access to account and billing information and some customers’ secrets, datasets and incomplete weights. Fireworks says it notified law enforcement and engaged Palo Alto Networks Unit 42 to assist with forensic work.
Notion separately said chat transcripts from its Notion AI Insights feature were in a dataset attackers may have accessed. The company said fewer than 1% of its workspaces were affected. That is Notion’s account of its exposure; it does not establish the full number of affected Fireworks customers or the scope of data held by other customers.
“For a limited set of accounts, we also discovered impact to secrets, datasets, and incomplete models.”
— Fireworks AI, November 9 security update
Scope and Data Fate Remain Unknown
The number of affected customers has not been disclosed. Fireworks has not identified the third-party cloud provider, explained how the internal credentials involved in the first access were obtained, or said whether attackers retained, copied, shared or used the data they accessed. Its description of the model weights as “incomplete” is also not defined.
Notion’s estimate applies to its own workspaces and does not answer how many other Fireworks customers may have been affected. Fireworks says its inquiry is ongoing and may produce additional findings, so the categories and scope it has reported could change. At present, the available information does not establish whether exposed data was misused.
Investigation and Customer Safeguards
Fireworks says it is continuing its investigation with help from Palo Alto Networks Unit 42 and has notified law enforcement. As the review proceeds, customers will need clearer information about which accounts and data were affected, what “incomplete” model weights contained, and whether any accessed secrets require rotation or other action.
The company has also disabled self-service dataset uploads and downloads used by Batch Inference and dataset-based training while it investigates. Fireworks has not provided a date for restoring those functions or announced a timetable for a final report. Further customer notices or security updates are the next likely source of detail.
Key Questions
What information does Fireworks AI say was accessed?
Fireworks reports access to resource metadata and billing information, and, for a limited set of accounts, secrets, datasets and a subset of incomplete model weights. It has not disclosed the number of affected customers.
Were payment card numbers exposed?
Fireworks says it does not store payment card numbers and that they were not affected. This is the company’s statement in its November 9 update.
Were Notion AI conversations involved?
Notion said chat transcripts from its Notion AI Insights feature were in a dataset attackers may have accessed. It said fewer than 1% of its workspaces were affected; the full scope of the Fireworks incident remains undisclosed.
What is Fireworks doing in response?
Fireworks says it engaged Palo Alto Networks Unit 42, notified law enforcement and disabled certain self-service dataset uploads and downloads while its investigation continues. It has not announced when those functions will return.
Source: ThorstenMeyerAI.com
