Fireworks AI Breach Widens to Expose Notion Customers' Chat Transcripts

TechCrunch reports that a cyberattack on AI inference provider Fireworks AI has widened, with hackers gaining access to chat transcripts belonging to Notion customers who used the Notion AI "Insights" feature, which ran on Fireworks' infrastructure. Fireworks had first disclosed unauthorized access to environment variables in a limited number of customer accounts, then confirmed attackers also reached its development machines through a third-party cloud provider, exposing partial model weights, datasets and billing information for some customers.

Notion said the incident affected fewer than 1% of its workspaces and that it has disabled Insights and stopped serving open-weight models through Fireworks. Fireworks attributed the activity to a financially motivated group it believes is seeking to sell unauthorized access to large language models, and TechCrunch noted this is the second breach to hit the company in as many months.

Why it matters: enterprises that build AI features on third-party inference and data platforms inherit that vendor's security posture, and this incident shows how a breach at the infrastructure layer can cascade into the model weights, datasets and customer data of everything built on top of it.