Fireworks AI says hackers reached customer datasets, partial model weights and billing data in a second intrusion

What happened? Fireworks AI, the AI inference and fine-tuning platform, said on 9 November 2026 that its investigation into an October security incident found more customer data was exposed than first disclosed. In an update to its security bulletin, the company said attackers reached account and billing information (billing and shipping addresses, invoice and payment history, balances and credit usage), and, for a limited set of accounts, secrets, datasets and a subset of incomplete model weights. Payment card numbers are not stored by Fireworks and were not affected.

Two routes in: Fireworks first detected unauthorised use of internal credentials on 6 October. The update adds a second path: a third-party cloud provider that hosted Fireworks R&D machines had a security incident on 23 October, and the next day the intruder used it to reach Fireworks development machines and run an unauthorised workload. Fireworks says it has seen no further successful access on its customer platform since 25 October, and no impact on Inference, Training SDK or Nexus, which are zero data retention by default. It has disabled self-serve dataset uploads and downloads while the investigation continues with Palo Alto Networks Unit 42, and notified law enforcement. It links the attack to a financially motivated "LLMjacking" group that resells stolen access to language models.

Why it matters: Bloomberg described it as the company's second known hack in a matter of weeks. Fireworks says it serves more than 10,000 customers, and a spokesperson said it is working with affected customers "including Notion, Cursor and others". The disclosure lands weeks after Fireworks reported passing $1B in annualised revenue and while it was reported to be eyeing a valuation of up to $40B, putting security at the centre of the inference-provider race.

Read more: Fireworks AI security bulletin, Bloomberg

Image: Fireworks AI