Incident sheet · /security
- System
- Security & Risk
- Filed
- Read
- 3 min · 585 words
- Severity
- Routine
Second Breach in Three Weeks Hits Fireworks AI via Cloud Provider
Fireworks AI hit by a second breach in three weeks after attackers reached its dev machines through a hacked cloud provider, exposing billing data and incomplete model weights.

Event log
- Fireworks disclosed a second breach on October 23 after a third-party cloud provider hack reached its development machines
- Attackers accessed billing data, resource metadata, and incomplete model weights and datasets; no payment card numbers were stored
- Fireworks says no further unauthorized access was detected as of October 25
- The group is financially motivated, selling stolen LLM tokens via 'LLMJacking' routing gateways
- Google Threat Intelligence and Mandiant report cloud AI charges from stolen API keys reaching $1 million
AI inference platform Fireworks disclosed its second security breach in under a month on October 23, after attackers compromised a third-party cloud provider and used it to run an unauthorized workload on the company's development machines.
The intrusion exposed customer billing information, resource metadata and incomplete model weights and datasets — but not payment card numbers, because Fireworks doesn't store them. The company detected the activity the same day and says it has found no evidence of successful follow-on access since October 25.
The San Francisco-area startup, founded in 2022, runs a high-performance inference platform that lets developers deploy open-source models — Mistral, Meta's Llama, Google's Gemma — without managing hardware or infrastructure themselves. That makes its API keys and customer credentials a prime target.
What happened, step by step?
The first incident surfaced in early October, when Fireworks notified law enforcement and customers that attackers had accessed environment variables holding API keys and tokens for a limited number of accounts.
"The evidence indicates links to a cybercriminal threat actor group, which is financially motivated and aimed at selling tokens from unauthorized access to Large Language Models (LLMs) to illicit routing gateways, known as 'LLMJacking,'" Fireworks said at the time.
The company revoked the compromised internal credentials, affected API keys, and any keys the attackers had created. Then came the second hit.
"A third-party cloud provider that hosted Fireworks R&D machines experienced a security incident on October 23," the company wrote in a Monday advisory. "On October 24, the unauthorized third party used it to reach Fireworks development machines and then run an unauthorized workload that led to additional customer data impact."
In that second breach, the threat actors accessed:
- Resource metadata: identifiers, display names, descriptions and creation times
- Billing data: billing and shipping addresses, invoice and payment history, account balances and credit usage
- Incomplete model weights and datasets belonging to some customers
"We detected this activity on the same day, and followed our incident response processes to take immediate remediation measures and close known critical vulnerabilities," the advisory continued. "As of October 25, we have not detected further evidence of successful unauthorized access by the threat actor on the Fireworks platform customers use."
How did Fireworks respond?
Since the attack, the company has revoked and replaced credentials and service accounts, blocked unauthorized network pathways, and isolated affected systems. It has also tightened access to data storage and analytics systems to shut down public or unauthorized reach into stored data.
Fireworks is now working with Unit 42, Palo Alto Networks' incident response division, to secure its systems against future attacks and to pin down the full scope and impact of the incident.
What does this change for AI platform buyers?
The incident lands in the middle of an escalating LLMjacking wave. One week before the disclosure, researchers at Google's Threat Intelligence Group and Mandiant reported that stolen API keys used for fraudulent model calls are driving record costs — with cloud AI charges for compromised organizations running as high as $1 million per incident.
For engineering and security leaders, the practical takeaway is straightforward: your inference vendor's secrets management is part of your own attack surface. Incomplete model weights and billing records leaking through a supplier's compromised R&D machines means third-party risk assessments for AI infrastructure now need the same rigor long applied to cloud and SaaS providers.
Fireworks hasn't disclosed how many customers were affected. Its investigation with Unit 42 continues.
via bleepingcomputer.com (Source)
More from Olivia Hart
Show full bio
News editor covering money, funding and market shifts around enterprise information technology and the leaders who run at CIO Monitor.
207 articles
The newsroom digest
One email a day with the stories that matter. No noise.
Cross-references
- Fireworks AI Breach Widened Beyond First Disclosure, Vendor Says
- Notion says Fireworks breach exposed customer AI chat transcripts
- Open-source AI platforms face a widening cybersecurity test
- Railway Breach Cost Hits $740,000 as Stolen AI Keys Resold
- Cursor Breach Exposed 5,000 Developers to Malicious Claude Code