🔍 Read the full analysis: Fireworks AI Breach Expands To Datasets And Model Weights on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
Fireworks AI says an October security incident affected account and billing information and, for a limited set of accounts, secrets, datasets and incomplete model weights. The company describes two routes of access, including an incident at a third-party cloud provider; Notion says transcripts from its AI Insights feature may have been in an accessed dataset.
Fireworks AI says an October security incident reached account and billing information and, for a limited set of accounts, secrets, datasets and incomplete model weights—a broader scope than the company disclosed in its initial October bulletin. In an update published November 9, the AI model platform said attackers used internal credentials and later accessed development machines through a third-party cloud provider; Notion separately said some of its AI Insights chat transcripts may have been in an affected dataset.
Fireworks’ October 9 bulletin said it had found misuse of internal credentials and access to environment variables in a limited number of customer accounts. At the time, the company said it had not detected unauthorized access to customer models, datasets, training data, inference traffic or personal information. Its November 9 update revised that account: account and billing information were affected, and certain accounts also had exposure involving secrets, datasets and incomplete models.
The company described two routes of access. It said an actor used internal Fireworks credentials in October to read environment variables from customers’ evaluation jobs. Separately, Fireworks reported that a third-party cloud provider hosting its research and development machines experienced a security incident on October 23. On October 24, an unauthorized party used that access to reach Fireworks development machines and run an unauthorized workload that led to additional customer-data impact. Fireworks said it detected that activity the same day.
Fireworks lists potentially affected information as resource metadata, billing details and, for certain customers, a subset of incomplete model weights and datasets. Billing information may include billing and shipping addresses, invoice and payment history, account balances and credit usage. The company says it does not store payment card numbers and that those numbers 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.
Fireworks AI breach expandsDatasets and model weights affected
A November update widened the disclosed scope of an October incident. Some customer accounts may have had secrets, datasets and incomplete model weights exposed.
What may be exposed
The reported scope includes different kinds of customer information and artifacts.
Account and billing details
May include addresses, invoice and payment history, account balances and credit usage.
Secrets and datasets
Some accounts may have had secrets and datasets accessed; dataset contents are not fully known.
Incomplete model weights
Fireworks reported exposure for a subset, but has not explained how much was accessible.
Two access routes
Fireworks described activity involving internal credentials and a third-party cloud provider.
An actor used Fireworks credentials to read environment variables from customer evaluation jobs.
After the provider incident, an unauthorized party accessed Fireworks development machines and ran a workload.
What customers should know
The impact depends on which accounts and files were involved.
- Payment card numbers were not stored or affected.
- Fireworks reported no detected impact to Inference, Training SDK or Nexus.
- No further successful access on the customer platform was detected after October 25.
- Total affected customers and accounts.
- Which data the datasets contained and how much model material was exposed.
- Whether accessed information was copied, retained or misused.
Response and follow-up
Fireworks says its investigation is ongoing, with forensic support from Palo Alto Networks Unit 42.
- Fireworks has notified law enforcement and is notifying affected customers.
- Self-service dataset uploads and downloads for Batch Inference and dataset-based training are disabled during the investigation.
- Notion said transcripts from its AI Insights feature may have been in an accessed dataset.
Risks to Customer Data and Models
The incident matters because companies use Fireworks to run and fine-tune open AI models, sometimes supplying proprietary datasets or model artifacts in the process. Exposure of datasets or model weights could affect customers’ confidential information and intellectual property, even if the materials are incomplete. The precise sensitivity and usefulness of the accessed files will depend on what each dataset contained and how much of each model was exposed.
The reported access to account and billing information also creates a separate concern for affected customers. Fireworks has not said how many accounts were involved, and the available information does not establish that exposed data was used for fraud or further attacks. The broader disclosure nevertheless changes the picture from the initial bulletin, which said no access to customer datasets or models had been detected.
For organizations using third-party AI infrastructure, the account also shows why the security of vendors and their cloud providers can matter alongside protections in the customer-facing product. Fireworks says its core inference and training products were not affected, but that statement does not resolve the exposure involving other accounts, datasets and development systems.
Popular right now
De'Longhi Magnifica Start Fully Automatic Espresso Machine with Milk Frother, Silver
Samsung 43-Inch M70H Series 4K Mini LED Smart TV (2026)
KitchenAid Artisan 5-Quart Tilt-Head Stand Mixer, Almond Cream
As an affiliate, we earn on qualifying purchases.
From October Alert to Expanded Disclosure
Fireworks said it identified the initial activity on October 6 and began containment. It began notifying affected customers on October 7, then published its first security bulletin on October 9 after rotating employee API keys. The company said it revoked additional user API keys and rotated service-account tokens on October 10.
The second route emerged later in the month. Fireworks said its cloud provider experienced a security incident on October 23 and that an unauthorized party reached development machines the following day. The company said it had not detected further successful unauthorized access by the actor on the Fireworks platform used by customers after October 25. It also said it notified law enforcement and brought in Palo Alto Networks Unit 42 to assist with forensic work.
Notion’s advisory describes how the expanded scope reached one customer. Notion said Fireworks first told it on November 2 about a small dataset that proved to be synthetic test data. On November 6, Fireworks notified the company that malicious actors may have accessed potential customer data, and on November 8 it confirmed the dataset involved. Notion said the dataset included chat transcripts from its Notion AI Insights feature and that fewer than 1% of workspaces were affected, according to its account.
“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 Use Still Unknown
Several key details remain undisclosed. Fireworks has not identified the third-party cloud provider, said how the internal credentials were first obtained, or given a total number of affected customers or accounts. The company has not explained what “incomplete” means for the model weights, including whether they were partial copies or artifacts from unfinished training runs.
It is also not clear what information the datasets contained beyond Notion’s description of its AI Insights transcripts, or whether the accessed material was copied, retained or used after access. Fireworks has not reported confirmed misuse of the information in the supplied update. Its investigation is ongoing, and the company says it may uncover further information.
Notion’s statement that fewer than 1% of its workspaces were affected is a company-reported figure. The available information does not specify the number of workspaces involved or establish whether other Fireworks customers had similar kinds of data in affected datasets.
Investigation and Customer Notifications
Fireworks says it is continuing its investigation, with forensic support from Palo Alto Networks Unit 42, and has notified law enforcement. It has also disabled self-service dataset uploads and downloads used by Batch Inference and dataset-based training while it investigates. The company says it detected no further successful unauthorized access on the customer platform after October 25, but has not announced when the investigation will conclude.
Customers may receive more specific information as Fireworks identifies affected accounts and files. Further updates will be needed to clarify the cloud provider incident, the number of customers affected, the nature of the model artifacts and datasets, and whether any accessed information was subsequently misused.
Key Questions
What information does Fireworks say was affected?
Fireworks says the incident reached account and billing information and, for a limited set of accounts, secrets, datasets and incomplete model weights. It also lists resource metadata as potentially accessible.
Were payment card numbers exposed?
Fireworks says it does not store payment card numbers and that they were not affected. The company says billing details such as addresses, invoices, payment history, account balances and credit usage may have been exposed.
What did Notion say about its users?
Notion said transcripts from its AI Insights feature were in a dataset attackers may have accessed. It reported that fewer than 1% of its workspaces were affected.
Which Fireworks products were not affected?
Fireworks says it has detected no impact to Inference, Training SDK or Nexus, which it describes as zero-data-retention by default. The company’s investigation is ongoing.
What remains unknown about the breach?
Fireworks has not named the cloud provider, disclosed the number of affected customers, explained how the internal credentials were obtained, or said what happened to the accessed data.
Source: ThorstenMeyerAI.com
