Fireworks AI breach exposes Notion AI chat transcripts, its second incident in two months

Fireworks AI, a platform developers use to access open AI models, disclosed that an attacker compromised its cloud provider and used that foothold to reach a Fireworks development system, TechCrunch reported on 12 November.

Fireworks says it has tens of thousands of customers, including Harvey and Cursor. Its security advisory says the attacker took customer billing information and some datasets.

Notion confirmed it was affected, putting the impact at about 1 per cent of its customers. The stolen material was chat transcripts from Notion AI users of the Insights feature, containing model inputs and outputs. Much of the data cannot be traced to a specific customer.

Notion says it disconnected its systems from Fireworks on 3 November, deleted data held in its own systems and cut all connectivity between the two companies.

Fireworks believes the attackers are financially motivated and aim to sell access to large language models obtained through breaches to unauthorised third parties. It is the company's second breach in two months, following an October incident.

Why it matters

The case shows that AI inference providers sit in the supply chain of many SaaS products, so a breach at one vendor can expose prompts and outputs from customers who never contracted with it directly.

It also illustrates a growing market for stolen model access, which turns credentials and API capacity into a monetisable target.

Indicops view

Indian enterprises should map which third-party model providers touch their data, demand breach-notification terms and retention limits, and avoid sending sensitive material through chains they cannot audit.