Fireworks AI breach exposes customer data in LLMjacking attack
Fireworks AI suffered two October breaches where hackers stole billing data and AI-specific information, with a third-party cloud attack enabling unauthorized workloads. The incident impacted Notion's systems and exemplifies the rapid financial fraud risk of LLMjacking attacks targeting AI infrastructure.

Fireworks AI confirmed a breach on October 6, 2026, where hackers accessed development environments via a third-party cloud provider to steal billing information and AI data. The AI infrastructure company, founded in 2022 with more than 10,000 customers, discovered unauthorized access at 9 a.m. PDT that day. Investigators identified a well-known financially motivated cybercriminal group as responsible for this first breach, which involved stealing billing information and AI-specific data from a limited number of customer accounts.
A second incident occurred on October 23 when an unnamed third-party cloud provider hosting Fireworks' R&D machines was attacked. The unauthorized party used that provider to reach Fireworks development machines and run an unauthorized workload on October 24. The breaches resulted in stolen internal Fireworks data, billing information, invoices, payment history, and AI-related datasets. Attackers accessed the development environment using compromised internal credentials, which allowed access to environment variables used in evaluation jobs. The breach affected resource metadata, billing information, and incomplete model weights and datasets for certain customers. Payment card numbers were not stored by Fireworks AI and were not impacted.
The company disclosed the security incident on October 10 and published an updated notice on the following Monday. Fireworks AI revoked the compromised internal credentials and rotated service account tokens, API keys, and keys created by the threat actor. It notified all affected customers and is working with Palo Alto Networks Unit 42 to investigate. The company has implemented security hardening procedures including credential revocation, system isolation, access control tightening, and improved logging and monitoring.
Notion's response and impact
Digital workspace company Notion confirmed some of its customers were impacted by the Fireworks breaches. Notion was notified of the incident on November 2 and promptly disconnected its systems from Fireworks, deleting all data. The company disclosed that the Fireworks AI incident impacted fewer than 1% of Notion workspaces. Notion AI chat transcripts used for the company's Insights feature were included in the impacted data, though the company stated this data was mostly unattributable. The Insights feature has been disabled as Notion investigates the breach.
LLMjacking context and financial impact
The breach exemplifies LLMjacking tactics where attackers hijack AI infrastructure for financial gain. Google Threat Intelligence Group and Mandiant published a blog about these attacks, warning that threat actors are increasingly targeting AI infrastructure across North America, South America, and Asia Pacific. The sectors affected include financial services, retail, technology, and transportation.
Mandiant reported responding to multiple LLMJacking incidents in 2026. The firm noted that threat actors were primarily focused on stealing tokens or credentials to establish footholds in target organizations' infrastructure. This tactic involves hijacking enterprise access to large language models for financial gain, with attackers maturing their approach by focusing on initial access before reselling unauthorized access to other threat actors. The legitimate owner is left to pay the bill.
| Region | Sector Impact | Financial Scale |
|---|---|---|
| North America | Financial Services, Retail, Technology, Transportation | Exceeding $100,000 in a single day |
| South America | Financial Services, Retail, Technology, Transportation | Cumulative charges reaching $1 million |
| Asia Pacific | Financial Services, Retail, Technology, Transportation | Cumulative charges reaching $1 million |
Mandiant highlighted the speed of the financial impact. "Unlike the gradual resource consumption associated with cryptojacking, LLMJacking can generate large spikes in fraudulent cloud charges in a matter of hours," the firm said. The incident at Fireworks is linked to a financially motivated cybercriminal threat actor group engaged in this type of LLMjacking activity.
Fireworks AI has advised affected customers to review impacted data and rotate credentials immediately if sensitive information was pasted during the attack window.





