AI infrastructure startup Fireworks AI has disclosed a new security breach after a third-party cloud provider hosting its R&D machines experienced a security incident on October 23. An unauthorized third party used it on October 24 to reach Fireworks’ development machines. The company detected the activity the same day and took immediate remediation measures, but said customer billing information, resource metadata, and incomplete model weights and datasets were accessed.
"A third-party cloud provider that hosted Fireworks R&D machines experienced a security incident on October 23," Fireworks said 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." The company said it detected the activity on the same day and followed its incident response processes to close known critical vulnerabilities.
The breach follows an earlier incident in early October, when attackers accessed environment variables containing API keys and tokens for a limited number of accounts. Fireworks notified law enforcement and customers at that time, and said evidence indicated links to a financially motivated cybercriminal group that aimed to sell tokens from unauthorized access to large language models to illicit routing gateways, known as LLMJacking.
Fireworks provides a high-performance inference platform that lets developers deploy and run open-source AI models without managing hardware or infrastructure. In the new breach, attackers accessed resource metadata, billing information including addresses and invoice history, and incomplete model weights and datasets for some customers. Fireworks said attackers could not access customer payment card numbers because it does not store them.
Since the attack, Fireworks says it has revoked and replaced credentials and service accounts, restricted unauthorized network pathways, isolated affected systems, and tightened access to data storage and analytics systems. The company is working with Palo Alto Networks' Unit 42 incident response division to further investigate and secure its systems.
One week ago, researchers at Google's Threat Intelligence Group and Mandiant reported that large language model hijacking attacks are driving record costs for compromised organizations, with cloud AI charges reaching as high as $1 million due to fraudulent calls using stolen API keys.
Analysis
Why This Matters
- Companies using Fireworks’ platform may face exposure of sensitive billing information and incomplete model data.
- The breach highlights the risks of third-party cloud providers in AI infrastructure supply chains.
- As LLM jacking attacks escalate, AI companies face growing financial and reputational costs from stolen API keys.
Background
Fireworks AI is an AI infrastructure startup that provides a platform for deploying and running open-source models. It disclosed an earlier breach in early October 2026 where attackers accessed environment variables containing API keys. The company linked that attack to a financially motivated group engaged in LLMJacking. The new incident, disclosed on November 10, involves a cloud provider compromise leading to further data access.
What to Watch
- The outcome of Unit 42’s investigation and any further details on the attackers’ use of stolen data.
- Whether Fireworks faces regulatory scrutiny or customer lawsuits over the back-to-back breaches.
- Indicators of whether the same threat group is responsible for both October incidents.
Checked
12 claims in this story checked against 1 source: 12 supported.
Show each claim
- SupportedFireworks disclosed a new breach after a cloud provider hack.BleepingComputer: “Fireworks AI discloses new breach after cloud provider hack.”
- SupportedAttackers accessed Fireworks’ development machines.BleepingComputer: “attackers accessed its development machines after breaching a third-party cloud provider.”
- SupportedAttackers accessed billing data and incomplete model weights and datasets.BleepingComputer: “the threat actors accessed resource metadata (e.g., Fireworks resource identifiers, display names, descriptions, creation time), billing information (e.g., billing and shipping addresses, invoice and payment history, account balance, and credit usage), and incomplete model weights and datasets for some customers.”
- SupportedThe quoted statement about the cloud provider’s October 23 incident is attributed to Fireworks in a Monday advisory.BleepingComputer: “"A third-party cloud provider that hosted Fireworks R&D machines experienced a security incident on October 23.”
- SupportedThe unauthorized third party used the provider on October 24 to reach Fireworks development machines and run a workload that led to additional customer data impact.BleepingComputer: “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,”
- SupportedFireworks detected the activity the same day, took immediate remediation measures and closed known critical vulnerabilities.BleepingComputer: “We detected this activity on the same day, and followed our incident response processes to take immediate remediation measures and close known critical vulnerabilities.”
- SupportedAn earlier incident in early October involved attackers accessing environment variables containing API keys and tokens for a limited number of accounts.BleepingComputer: “The company notified law enforcement and customers of another security incident in early October after attackers accessed environment variables containing API keys and tokens stored as secrets for a limited number of accounts.”
- SupportedFireworks notified law enforcement and customers about the earlier incident.BleepingComputer: “The company notified law enforcement and customers of another security incident in early October”
- SupportedFireworks said evidence indicated links to a financially motivated group aiming to sell tokens obtained through unauthorized LLM access to illicit routing gateways, a practice known as LLMJacking.BleepingComputer: “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'”
- SupportedAttackers could not access customer payment card numbers because Fireworks does not store them.BleepingComputer: “the company said the attackers couldn't access customer payment card numbers because Fireworks doesn't store them.”
- SupportedSince the attack, Fireworks says it revoked and replaced credentials and service accounts, restricted unauthorized network pathways, isolated affected systems, tightened access to data storage and analytics systems, and is working with Unit 42 to investigate and secure its systems.BleepingComputer: “Since the attack, Fireworks revoked and replaced credentials and service accounts, restricted unauthorized network pathways, isolated affected systems, and tightened access to data storage and analytics systems to block public or unauthorized access to stored data. The company is now working with Palo Alto Networks' Unit 42 incident response division to secure its systems against future attacks and further investigate the incident's full scope and impact.”
- SupportedOne week ago, Google’s Threat Intelligence Group and Mandiant researchers reported cloud AI charges reaching as high as $1 million due to fraudulent calls using stolen API keys.BleepingComputer: “One week ago, researchers at Google's Threat Intelligence Group (GTIG) and Mandiant reported that large language model hijacking attacks (also known as LLMjacking) are driving record costs for compromised organizations , with cloud AI charges reaching as high as $1 million due to fraudulent calls using stolen API keys.”
