Fireworks AI Hack Exposes Notion Customers’ Chat Transcripts

A cyberattack targeting artificial intelligence infrastructure provider Fireworks AI has exposed sensitive customer information, including chat transcripts belonging to users of popular productivity platform Notion. The incident highlights the security risks companies face when relying on third-party technology providers to power their AI services.

Fireworks AI confirmed that an attacker gained unauthorized access to one of its development systems after breaching its cloud service provider. This access allowed the hackers to obtain some customer information stored within the company’s environment.

According to Fireworks AI’s latest security advisory, the stolen information included customer billing details and certain datasets. The company provides developers with access to open artificial intelligence models, allowing businesses to integrate AI capabilities into their own products and services.

Fireworks AI serves tens of thousands of customers worldwide. Its customer base includes prominent technology companies such as legal technology startup Harvey and AI coding platform Cursor.

However, one of the most notable companies affected by the incident is Notion, the online workspace and document management platform. Notion confirmed that approximately 1% of its customers were affected by the security breach.

The company said Fireworks initially informed it about the security incident before later confirming that attackers had accessed and stolen some Notion customer data.

On November 3, Notion took steps to protect its systems by disconnecting its infrastructure from Fireworks AI. It also deleted the data stored within the affected systems and terminated all connections between the two companies.

Notion has now confirmed that the stolen information included chat transcripts from customers who used its Notion AI service through the platform’s “Insights” feature.

These transcripts contain information entered by users and responses generated by the AI model. Notion explained that any input provided to the model or output generated by it that was recorded in a chat would be included in the corresponding transcript.

However, the company noted that much of the exposed information could not be linked to specific individuals. As a result, Notion was unable to determine which customers were associated with certain portions of the stolen data.

The incident demonstrates how security problems at one technology company can create risks for businesses that depend on its services. Even when a company maintains its own security systems, information shared with external infrastructure providers can become vulnerable through third-party breaches.

Fireworks AI believes the attackers were primarily motivated by financial gain. According to the company, their activities appear to be focused on obtaining and selling access tokens associated with large language models (LLMs).

These tokens can allow unauthorized third parties to use AI models without paying for the resources themselves. By selling stolen access, attackers can enable others to consume AI computing resources using compromised accounts or credentials.

The latest incident also raises concerns about the security of AI infrastructure providers as more businesses integrate large language models into their everyday operations.

For Fireworks AI, this is not the first security issue in recent months. The company experienced another security incident in October, making this its second reported breach in as many months.

With AI adoption continuing to grow, the incident serves as another reminder that protecting customer information requires strong security practices across the entire technology supply chain, not just within individual companies.

Also read : Plurai Raises $13M to Make AI Agents More Reliable