Fireworks AI says breach exposed customer data, secrets and model weights
Original title: Fireworks AI Hit by LLMJacking Cyberattack Exposing Customer Data and AI Model Weights
Stolen credentials were used on October 6, and a later cloud-provider incident exposed more customer data.
Fireworks AI said financially motivated criminals stole internal credentials and pursued LLM-jacking by reselling model capacity through illegal routing gateways. An October 6 intrusion and a second intrusion on October 24, after an incident at a third-party cloud provider, exposed account information, billing records, secrets, datasets and incomplete model weights for some customers. Payment card numbers were not stored. Inference, Training SDK and Nexus showed no detected impact. Unit 42 is assisting, and the findings remain provisional.
Unconfirmed: Fireworks says the findings remain provisional and has not named the responsible group.
- Unauthorized credential use was detected on October 6 at 9 a.m. PDT.
- Attackers accessed research and development machines on October 24.
- No successful customer-platform access has been detected since October 25.
- Employee API keys were rotated and additional tokens revoked in October.
- Affected customers should replace revoked keys and rotate third-party secrets.
Impact: Customer data, secrets, datasets and incomplete weights exposed
| Product | Affected |
|---|---|
| Fireworks AI platform | — |
- 1.Replace revoked keys and rotate third-party secrets with the issuing providers.
- 2.Review logs, users, jobs, datasets and keys for unfamiliar activity.
- 3.Update applications that still rely on revoked credentials.
- Victims
- Fireworks AI
- Organizations
- Palo Alto Networks
From the original article
2 min read at gbhackers.com
Fireworks AI has disclosed a wider security breach involving customer account information, billing records, secrets, datasets, and incomplete model weights. Its November 9 update links the activity to financially motivated criminals pursuing LLM-jacking, which involves stealing access to large language models and reselling their capacity through illegal routing gateways. The investigation remains ongoing.
By Divya · gbhackers.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.