Fireworks AI says hackers accessed customer datasets and model weights
AINews

Fireworks AI says hackers accessed customer datasets and model weights

Fireworks AI said on November 9, 2026 that hackers accessed some customers' datasets, incomplete model weights and billing information, expanding an October security disclosure. The company says payment cards were not stored or impacted, and it has not detected impact to Inference, Training SDK or Nexus. Its findings rest on its investigation to date.

Fireworks AI’s November 9, 2026 bulletin says attackers reached billing data and, for some accounts, secrets, datasets and incomplete model weights; Inference, Training SDK and Nexus were not impacted.

Close-up of dual computer monitors with green coding interfaces in a dark room, an illustrative stock photograph unrelated to Fireworks AI
Photo by Tima Miroshnichenko on Pexels

Fireworks AI said on November 9, 2026 that hackers accessed some customers' datasets, incomplete model weights and billing information, expanding an October security disclosure. The company says payment cards were not stored or impacted, and it has not detected impact to Inference, Training SDK or Nexus.

Forbes estimates Fireworks AI has about 10,000 enterprise customers, $250 million in 2025 revenue, and a $17.5 billion valuation.

In that security bulletin, Fireworks updated its October 9, 2026 notice. After the earlier incident, it said, investigators found impacts to account and billing information, and for a limited set of accounts also to secrets, datasets and incomplete models. A third-party cloud provider that hosted Fireworks research-and-development machines had a security incident on October 23; on October 24 an unauthorized party used that path to reach Fireworks development machines and run an unauthorized workload that led to further customer-data impact, according to the company. Fireworks said it detected that activity the same day, and that as of October 25 it had not detected further successful unauthorized access on the platform customers use. Evidence, it said, links the activity to a financially motivated group aimed at selling unauthorized LLM access through illicit routing gateways, a pattern known as LLMJacking. Fireworks notified law enforcement and is working with Palo Alto Networks Unit 42. Billing data in scope included addresses, invoice and payment history, account balances and credit usage; Fireworks does not store payment card numbers. Self-serve dataset uploads and downloads used by Batch Inference and dataset-based training are disabled while the investigation continues, the bulletin says. Training SDK remains available with zero data retention, and Fireworks Lab-based training operates in an isolated environment with added network and storage protection, per the company.

Bloomberg described the November disclosure as Fireworks' second known hack in weeks. The Information reported that Fireworks had told large customers, including legal AI company Harvey, about a more serious recent hack. Forbes spoke with two customers who said they had already substantially moved business off Fireworks' systems; one described a series of "very bad hacks" and criticized Fireworks' customer communications. Two other customers told Forbes they were not aware of any Fireworks hacks. Forbes separately estimates Fireworks has about 10,000 enterprise customers, booked about $250 million in 2025 revenue, and reached a $17.5 billion valuation, figures it also used in an October 6, 2026 feature. Builders who keep sensitive training data and unfinished model assets on third-party platforms face the same trade-offs Codema covers in owning your code and data and in MCP security best practices. Fireworks' findings remain based on its investigation to date, the company said, and the inquiry is ongoing.

FAQ

What did Fireworks say was accessed?

In a November 9, 2026 bulletin updating an October 9 notice, Fireworks said investigators found impacts to account and billing information, and for a limited set of accounts also to secrets, datasets and incomplete models. Billing data in scope included addresses, invoice and payment history, account balances and credit usage.

Were payment cards stolen? Were Inference, Training SDK or Nexus affected?

Fireworks said payment cards were not stored or impacted, and that it has not detected impact to Inference, Training SDK or Nexus.

What does Fireworks say happened?

A third-party cloud provider that hosted Fireworks research-and-development machines had a security incident on October 23, and on October 24 an unauthorized party used that path to reach Fireworks development machines and run an unauthorized workload, the company said. Fireworks said it detected that activity the same day, and as of October 25 had not detected further successful unauthorized access on the platform customers use. It said evidence links the activity to a financially motivated group aimed at selling unauthorized LLM access, a pattern known as LLMJacking.

Sources

  1. Fireworks AI, Update on October 2026 security incident (last updated Nov. 9, 2026)fireworks.ai
  2. The Information, New Fireworks Hack Threatens $1 Billion Business (Nov. 9, 2026)theinformation.com
  3. Bloomberg, Nvidia-Backed Fireworks AI Notifies Customers of Its Second Hack in Weeks (Nov. 9, 2026)bloomberg.com
  4. Forbes, Fireworks Cyberattack Exposes Customer Data And AI Models As Fallout Widens (Nov. 9, 2026, Anna Tong and Rashi Shrivastava)forbes.com
  5. Forbes, AI Costs Too Much. The Reckoning Is Here. (Oct. 6, 2026, Rashi Shrivastava)forbes.com

Topics

John F.AI desk

John F. is a Codema.ai house byline: a pen name used by our editorial desk, not one individual journalist. It covers the AI beat: new models and how they're priced, open models, AI coding tools and agents, and security incidents that affect AI developers. Pieces under this name are written with AI assistance and reviewed by a human editor before they're published.

All stories by John F.

More in AI

See all
  1. Helsing unveils Clarity, European rival to Palantir's Maven

    Kelly B.

  2. DeepMind's AlphaGenome 2 predicts DNA variant cell effects

    Kelly B.

  3. Gemini Decisions API: 1,000 questions at $0.10 per 1M input

    John F.

Matching resources