Skip to content
Stack Courier SC markSCSTACKCOURIER
Fireworks AI breach exposes Notion chat logs and billing data
Depth
1 outlet · newest source 92h ago
Foundation models

Fireworks AI breach exposes Notion chat logs and billing data

Hackers accessed Fireworks systems, stealing customer datasets and billing info, marking the firm's second incident in two months.

Key points

  • Attackers breached Fireworks' cloud provider to reach development systems.
  • Notion confirmed exposure of AI chat transcripts for some users.
  • Fireworks suspects financial motives tied to selling LLM tokens.

TechCrunch reported that Fireworks AI confirmed a cyberattack exposed customer data after hackers breached its cloud provider. The intrusion granted access to a development system, allowing theft of billing information and datasets from clients including Notion.Hackers broke into Fireworks AI's systems and stole customer data, including billing info and chat logs from Notion users.An intrusion into Fireworks' cloud infrastructure allowed attackers to exfiltrate billing data and datasets, prompting Notion to sever integration links.

Evidence: reportedCyberattack on Fireworks AI allowed hackers to access Notion customers’ chat transcripts

aifleet-techcrunch-all

The facts

The company believes the attackers are financially motivated, aiming to sell access tokens for large language models to unauthorized third parties who can burn through AI model usage without using their own resources.

Who it affects

Notion confirmed it is an affected customer, stating the breach impacted roughly one percent of its user base. Other known Fireworks customers include legal tech firm Harvey and AI company Cursor, though specific impact details for them were not provided in the report.

What is still unknown

It remains unclear exactly which datasets beyond billing were fully compromised across all customers. The material does not specify if other named clients like Harvey or Cursor suffered data loss, nor does it detail the full scope of the stolen information beyond Notion's subset.

The takeaway

The incident highlights risks in relying on third-party AI infrastructure. Fireworks noted this is its second breach in as many months, following an October incident, suggesting ongoing security challenges for platforms that host or facilitate access to open AI models for developers.

What changes for users

Notion users who utilized the AI Insights feature may have had their chat transcripts exposed. These transcripts contain inputs and outputs written to the chat, though Notion noted much of the data was unattributable. Notion disconnected from Fireworks on November 3 and deleted stored data, stopping all connectivity between the two systems.

Evidence: reportedCyberattack on Fireworks AI allowed hackers to access Notion customers’ chat transcripts

aifleet-techcrunch-all

Why it matters

  • Data science teams using Fireworks should verify if their datasets were accessed, as billing and model inputs were compromised.
  • The financial motive of selling LLM tokens indicates attackers aim to sell access to unauthorized third parties for AI usage.
  • Notion's disconnection from Fireworks on November 3 shows how downstream customers respond to compromised vendor integrations.
  • This is the second breach in two months for Fireworks, raising questions about the stability of third-party AI infrastructure for enterprise clients.

Topics: AI security, data breach, Fireworks AI, Notion, LLM

Updated

Then / now / next

Then

Fireworks operated as a platform for accessing open AI models without recent publicized security incidents.

Now

Hackers have accessed development systems, stealing billing data and customer datasets including Notion chat logs.

Next

Teams should monitor for further disclosures regarding the scope of data loss across other Fireworks customers.

Story timeline

  1. Then Fireworks operated as a platform for accessing open AI models without recent publicized security incidents.
  2. 12 Nov 2026 Cyberattack on Fireworks AI allowed hackers to access Notion customers’ chat transcripts
  3. Now Hackers have accessed development systems, stealing billing data and customer datasets including Notion chat logs.
  4. Next Teams should monitor for further disclosures regarding the scope of data loss across other Fireworks customers.

The strongest doubt

Counterpoint from the sources

Notion noted that much of the stolen data was unattributable, meaning it could not be determined from which specific customer the information was obtained.

Ask the article

Local keyword search of this article; no generated answers.

Did this article give you the evidence you needed?

No vote saved on this device.
Your choice stays on this device. No public tally.

Questions readers ask

Did the breach affect all of Notion's users?

No, Notion stated the incident affected approximately one percent of its customer base.

What specific data was stolen from Notion?

Hackers accessed chat transcripts from the Notion AI Insights feature, including inputs and outputs from the model.

Is this the first breach for Fireworks AI?

No, this is the second security incident affecting the company in as many months, following one in October.

Connected stories

This storyAI-Enhanced Infostealer Volta Uses LLMs to Profile VictimsAI-Enhanced Infostealer Volta UsInternal PDF QA tests reveal hidden prompt injection risks in financial data processingInternal PDF QA tests reveal hidCrowdStrike exec says AI hacks use known paths, not new exploitsCrowdStrike exec says AI hacks uChainalysis Confirms Credential Theft for All Users After Multi-Front AttackChainalysis Confirms Credential

Reading receipt

  • Attackers breached Fireworks' cloud provider to reach development systems.
  • Notion confirmed exposure of AI chat transcripts for some users.
  • Fireworks suspects financial motives tied to selling LLM tokens.

Still open: What specific datasets were compromised for customers other than Notion?

Still open: Did the October breach share similar vectors or attackers with this incident?

Time spent: 0:00 · Depth: Full

Sources

  1. TechCrunch (all)

Related stories