Fireworks.ai Update on October 2026 security incident (fireworks.ai)

Fireworks.ai has released an update regarding a security incident first reported on October 9, 2026, revealing that unauthorized access extended beyond stored secrets to include account and billing information, as well as data on incomplete models and datasets for a limited number of users. The company, in collaboration with Palo Alto Networks Unit 42, is continuing its investigation after evidence linked the breach to a cybercriminal group engaging in "LLMJacking," which focuses on stealing tokens from Large Language Models (LLMs) for monetary gain. A subsequent breach on October 23, involving a third-party cloud provider, further compromised development machines and customer data, but Fireworks promptly detected this activity and implemented response measures. The incident underscores significant challenges in securing AI systems and highlights vulnerabilities in third-party cloud infrastructures. Fireworks has proactively fortified its security by revoking credentials, isolating affected systems, and tightening data access protocols. Although no customer-facing services like Inference and Training SDK were impacted, the company is taking extensive steps to enhance its cybersecurity posture, including disabling certain dataset functionalities during the investigation. Ongoing monitoring and audits aim to mitigate further risks as Fireworks works transparently with affected customers on response actions.