Safety

AI-generated text

Fireworks AI cloud breach exposed Notion chat transcripts and customer billing data

A cloud breach at Fireworks AI allowed attackers to access some customers’ billing records and datasets, the company says.

Fireworks AI cloud breach exposed Notion chat transcripts and customer billing data

Fireworks AI has confirmed a cyberattack that exposed data belonging to some of its customers. Fireworks AI operates a platform that developers use to access and run open AI models; the company says an attacker compromised its cloud provider and used that access to reach a Fireworks development system, enabling the theft of certain customer information.

Types of data taken

In its latest security advisory, Fireworks AI stated that the attackers obtained customer billing information and some datasets. The company says it serves tens of thousands of customers worldwide, and named customers include the legal tech company Harvey and the AI company Cursor.

Notion affected and immediate response

At least one confirmed affected customer is Notion, the online document editor and note-taking service. Notion disclosed that the breach affected roughly 1% of its customers. According to Notion, Fireworks alerted them to the security incident and later confirmed that the attackers had stolen some Notion customer data.

On November 3, Notion disconnected its systems from Fireworks, deleted data stored in its systems related to the integration, and stopped all connectivity between the two services.

What was in the stolen data

Notion confirmed this week that the attackers took chat transcripts of Notion AI customers who used the "Insights" feature. Notion said those transcripts include any input to or output from the model that was written into the chat; however, much of the data was unattributable and it could not be determined which customer some entries came from.

Motive and how attackers profit

Fireworks believes the attackers were financially motivated and focused on selling tokens obtained through breached access to large language models (LLMs). Such schemes involve selling access to unauthorized third parties, who can then consume AI model usage without using their own resources.

Repeated incidents and supply-chain risk

This is the second breach affecting Fireworks in recent months, following a security incident in October. The events underscore how reliant tech companies are on third-party providers and how an attack on one vendor can cascade to impact other companies and their users.

Both Fireworks and affected customers have said they will provide further updates as investigations continue and more details about scope and impact are determined.