Fireworks AI breach leaked Notion AI chat transcripts

Attackers broke into Fireworks AI and some Notion AI conversations leaked. Here is what to rotate and review if you use AI providers.

Fireworks AI breach: AI vendor breach exposed chats

TL;DR: Attackers broke into Fireworks AI, a company that hosts AI models, and chat logs from Notion AI in a few Notion workspaces were exposed. If you rely on an AI provider, rotate the keys it holds and review what you send it.

What happened

BleepingComputer reported that Notion, the workspace platform, disclosed an intrusion at Fireworks AI, a supplier behind its AI features. Notion said some conversations with its AI assistant were exposed, but only for a handful of workspaces. Other outlets are covering a wider breach at Fireworks AI that touched more customer data, so details may still change.

The lesson is about supply chains. When a vendor handles your data, a breach on its side becomes your incident, even though your own systems were untouched.

Am I affected?

Notion is the customer named so far. If you use Notion AI, watch for notices from Notion about your workspace. If you use Fireworks AI directly, or any third-party AI service, treat the API keys you gave it as possibly exposed: rotate them, check your usage and billing for activity you do not recognise, and review which prompts and documents your teams send to that provider.

How to check with Stackurity

Be clear about the limits: a breach at a hosted service is not a vulnerability in your software, so Stackurity does not detect it. It can help with the part you control. Stackurity finds keys, tokens and passwords written in files on your servers and repositories and lists where they are, never their values, so you know what to rotate and where old copies remain. See the remediation documentation for how fixes are tracked. Stackurity Cloud has a 14-day trial: start a free trial.