Fireworks AI says two October breaches exposed billing data and datasets

Fireworks AI says two October breaches exposed billing data and datasets

10 November 2026: Fireworks AI, the AI inference company, says it suffered two separate breaches in October, not one. According to The Record, an updated notice published on Monday 9 November says investigators found a financially motivated criminal group stole account and billing information and secrets, and that a second intrusion on 24 October exposed invoices, payment history and datasets.

Not yet independently verified. The widened findings (the second incident, the stolen billing data and datasets, and the role of Unit 42) rest on The Record's account of Fireworks' updated notice; the version of the bulletin we could read is dated 10 October and does not contain them. Notion's statements are quoted as The Record reports them. We will update this when it can be confirmed, and remove this note.

The 6 October incident, in Fireworks' own words

Fireworks' security bulletin, in the version we could read, describes the first incident. Fireworks identified unauthorised use of internal credentials on 6 October and began containment that morning, Pacific time. An unauthorised third party accessed environment variables used in evaluation jobs in a limited number of customer accounts, and Fireworks believes API keys and tokens stored as secrets were captured.

At that stage Fireworks said it had found no evidence of access to customer models, datasets, training data, personal information or inference traffic. It added that it does not store prompts or outputs by default. It began notifying customers on 7 October, rotated employee API keys on 9 October and revoked more user API keys and service-account tokens on 10 October.

The bulletin told affected customers to replace revoked keys and to rotate third-party credentials with the providers that issued them, since removing a stored secret does not invalidate the credential behind it.

What the updated notice adds

The Record reports that Fireworks hired Palo Alto Networks Unit 42, which found that a well-known, financially motivated group carried out the first attack and stole account and billing information as well as secrets. Fireworks said the incident affected a "limited number of customer accounts", all of which were notified.

The same notice describes a second incident. On 23 October a third-party cloud provider hosting Fireworks' research and development machines was attacked. On 24 October, Fireworks says, the unauthorised party used that provider to reach development machines and run an unauthorised workload, causing further customer data impact. The Record lists internal data, billing information, invoices, payment history, datasets and other AI-related data. Affected customers have been notified, and Fireworks is working with Palo Alto Networks on hardening and on revoking compromised credentials.

The Record describes Fireworks as based in San Mateo, founded in 2022, with more than 10,000 customers.

Notion's exposure

The Record also reports a statement from Notion, which said Fireworks notified it of the incident on 2 November, and that on 6 November Fireworks told it that malicious actors may have accessed potential customer data. Notion used Fireworks for classification jobs behind its Insights feature and as a general inference provider for open-weight models inside its product. It has disabled Insights while it investigates and no longer serves open-weight models through Fireworks.

Where Google's LLMJacking warning fits

Fireworks said its breaches belong to a wider run of attacks that incident responders at Google have described. The Record notes that the Google Threat Intelligence Group and Mandiant published research last week on LLMJacking, in which attackers hijack an enterprise's access to large language models. Mandiant reported fraudulent charges above $100,000 in a single day, and cumulative charges reaching $1 million at some organisations before access was revoked.

Sources

• Fireworks AI: Fireworks October 2026 security incident, security bulletin (version last updated 10 October 2026), 10 October 2026
• The Record: AI company Fireworks says hackers stole data, billing info during October breaches, 10 November 2026, 10 November 2026