TechCapsules

AI-summarised brief · reviewed before publication

Fireworks AI announced that a security breach, initially detected in early October, had expanded to compromise customer billing data, secrets, and incomplete model weights for a limited set of accounts. The intrusion began on October 6 when attackers accessed internal credentials, later exploiting a third‑party cloud provider to reach development machines on October 23–24. Fireworks revoked compromised credentials, rotated tokens, and isolated affected systems. No customer models, training data, or personal information were accessed, and payment card numbers were not stored by the company.