🔍 Read the full analysis: Check Your AI Data Practices Amid The Fireworks AI Breach on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Fireworks AI’s November 9 update says an incident first disclosed in October reached account and billing information and, for a limited set of accounts, secrets, datasets and incomplete model weights. Notion says chat transcripts from its AI Insights feature may have been in an accessed dataset, affecting fewer than 1% of workspaces. The number of affected customers and what happened to accessed data remain unknown.
Fireworks AI said on November 9 that an October security incident reached further than the company initially reported, exposing account and billing information and, for a limited set of accounts, secrets, datasets and incomplete model weights. The update changes the scope of a breach at a platform used by companies to run and fine-tune AI models, though Fireworks has not disclosed how many customers were affected.
In an October 9 bulletin, Fireworks said someone had used internal credentials to read environment variables from a limited number of customer evaluation jobs. It said then that it had not detected unauthorized access to customer models, datasets, training data, inference traffic or personal information. Its November 9 update described additional impacts discovered during the investigation: account and billing information, plus secrets, datasets and incomplete models for a limited set of accounts.
Fireworks describes the exposed information as resource metadata, billing and shipping addresses, invoice and payment history, account balances and credit usage. For some customers, it says, a subset of incomplete model weights and datasets was accessed. The company has not explained what “incomplete” means in this case or identified the customers whose data was involved.
The company describes two routes of access. In October, internal Fireworks credentials were used to read environment variables from customer evaluation jobs. Separately, a third-party cloud provider hosting Fireworks research and development machines experienced a security incident on October 23. Fireworks says an unauthorized party reached its development machines the next day and ran an unauthorized workload that caused additional customer-data impact. The company says it detected that activity on October 24.
Check Your AI Data Practices Amid the Fireworks AI Breach
Fireworks AI’s November update says an October incident reached more customer information than first reported. The full scope and what happened to accessed data remain unknown.
What the incident reached
The reported data varies by route and account; the total number of affected Fireworks customers is undisclosed.
Billing and resource details
Fireworks lists addresses, invoices, payment history, account balances, credit usage and resource metadata.
Secrets and model materials
Secrets, datasets and incomplete model weights were accessed for a limited set of accounts.
AI Insights transcripts
Notion said chat transcripts were in an accessed dataset; fewer than 1% of its workspaces were affected.
How findings expanded
Fireworks described internal credentials and a later incident involving a third-party cloud provider’s development machines.
What remains unclear
The investigation has not resolved the breadth or consequences of the exposure.
- Total number of affected Fireworks customers and accounts
- How the internal credentials were obtained
- Whether accessed data was copied, retained, shared or used
- What “incomplete” model weights contained
- Which cloud provider was involved
Customer review checklist
Check the information and access connected to your AI services.
Review Fireworks’ account-specific updates for details relevant to your organization.
Identify credentials, datasets and model artifacts supplied to affected services.
Determine whether your internal incident response or customer notification process applies.
Risks for Companies Using Fireworks
The incident matters because Fireworks is infrastructure for other companies’ AI work. Customers may send the platform datasets, credentials or model artifacts to train and operate systems. Exposure at a service provider can therefore involve information tied to customers’ products and operations, not just the provider’s own records.
Incomplete model weights may still be sensitive: they are numerical representations produced during model training, and the data they contain or reveal can matter to the company that trained the model. Fireworks has not said what the accessed weights contained or whether they could be used to reconstruct customer information. The reported access should not be treated as proof that any such reconstruction or misuse occurred.
The update also gives customers a reason to review what data and credentials they supplied to AI vendors, where those materials were stored and what access they granted. Fireworks says it has not detected impact to its Inference, Training SDK and Nexus products, which it describes as zero-data-retention by default. That statement is specific to those products and does not negate the reported access to other data categories.
Popular right now
KitchenAid Artisan 5-Quart Tilt-Head Stand Mixer, Almond Cream
Dyson V15 Detect Cordless Vacuum
Vitamix VX1 Professional-Grade Blender, 64 oz, Black
As an affiliate, we earn on qualifying purchases.
From Initial Alert to Broader Findings
Fireworks says it identified activity on October 6 and began containment, then started notifying affected customers on October 7. On October 9, it published its first bulletin and said it had rotated employee API keys. It reported revoking more user API keys and rotating service-account tokens on October 10.
The second reported route emerged later in October. Fireworks says the cloud provider hosting its research and development machines experienced an incident on October 23, followed by unauthorized access to Fireworks development machines on October 24. The company says it detected the activity that day and has not detected further successful unauthorized access by the threat actor on the customer-facing platform since October 25.
Notion separately said Fireworks contacted it on November 2 about a small dataset initially understood to be synthetic test data. On November 6, Fireworks told Notion that malicious actors may have accessed potential customer data, and the dataset was confirmed on November 8. Notion said chat transcripts from its Notion AI Insights feature were in the dataset and that fewer than 1% of its workspaces were affected. That figure is Notion’s statement about its workspaces, not a count of all Fireworks customers.
“For a limited set of accounts, we also discovered impact to secrets, datasets, and incomplete models.”
— Fireworks AI, November 9 security update
Customer Scope and Data Use
Several key details remain undisclosed. Fireworks has not named the third-party cloud provider, given a total number of affected customers or explained how the internal credentials used in the first route were obtained. It also has not said whether accessed data was copied, retained, shared or used after access.
The company says the investigation is ongoing and that it has not detected further successful access to the customer-facing platform since October 25. That is a statement about what its investigation has detected so far, not confirmation of what happened to every accessed item. Fireworks has not publicly identified the group responsible; the supplied account says it referred to evidence of links to a cybercriminal, but provides no further detail.
Notion’s fewer-than-1% figure applies to its workspaces and does not establish the overall scale of the Fireworks incident. The available information also does not specify what the Notion transcripts contained or whether any data was published or otherwise misused.
Investigation and Customer Safeguards
Fireworks says it notified law enforcement and engaged Palo Alto Networks Unit 42 to conduct forensic work. It has also disabled self-service dataset uploads and downloads used by Batch Inference and dataset-based training while its investigation continues. The company says it may disclose additional information as the review progresses.
Customers using Fireworks can check the company’s notices for account-specific details and review credentials, datasets and model artifacts associated with affected services. They may also need to assess whether the information involved calls for internal incident response or customer notification. The source material does not establish that any particular customer has an obligation to take those steps.
The next major information to watch for is a more precise account of the affected accounts and data, the cloud provider incident, and any evidence about what happened after unauthorized access. Until Fireworks or affected customers provide those details, the breadth and consequences of the exposure remain partly unresolved.
Key Questions
What did the Fireworks AI incident expose?
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, billing and shipping addresses, invoice and payment history, account balances and credit usage.
Were payment card numbers exposed?
Fireworks says it does not store payment card numbers and that they were not impacted. This is the company’s statement in its November 9 update.
Were Notion AI conversations involved?
Notion said chat transcripts from its AI Insights feature were in a dataset that attackers may have accessed. It said fewer than 1% of its workspaces were affected; the report does not specify the contents of the transcripts or indicate that they were misused.
How many customers were affected?
Fireworks has not provided a total number of affected customers. Notion’s figure applies to its own workspaces and should not be read as the overall count for the Fireworks incident.
What is Fireworks doing now?
Fireworks says it has engaged Palo Alto Networks Unit 42, notified law enforcement and disabled certain self-service dataset uploads and downloads while it investigates. It says it has not detected further successful unauthorized access to the customer-facing platform since October 25.
Source: ThorstenMeyerAI.com
