Fireworks AI

Cursor released its investigation into an October security incident in which attackers used leaked Fireworks keys to steal data from a shared storage bucket, then leveraged internal credentials to inject a tampered Claude Code extension into the Cursor marketplace. The malicious build stayed live for about 11 hours and affected up to 5,000 users; fewer than 0.05% of non-privacy-mode users had agent trace data exposed, and Cursor has notified affected customers and rotated credentials.

Fireworks AI confirmed that an attacker breached its cloud provider to compromise a development system, exposing some customer billing information and datasets. Notion disclosed that roughly 1% of its customers were affected, disconnected its systems from Fireworks on November 3, and deleted stored data. It marks the second security incident for Fireworks in two months.

Notion disclosed a security incident at its vendor Fireworks AI that may have affected fewer than 1% of Notion workspaces, with some unattributable Notion AI chat transcripts potentially exposed. The chats carried no user or workspace labels and contained no tool call results, and no other workspace data was affected. The incident highlights third-party supply-chain risk propagating through AI infrastructure providers.

The Information reports that Fireworks, one of the highest-valued inference providers, was breached on October 24 via development machines hosted by a third-party cloud provider. The intrusion exposed customer billing records and, for some accounts, datasets and incomplete model weights. Customers including Harvey are now weighing a move off the platform, threatening Fireworks' billion-dollar business.