Overview
A breach involving AI infrastructure provider Fireworks AI has exposed some Notion AI chat transcripts, demonstrating how a security incident at an external model provider can affect users of a separate workplace application.
The story has developed beyond the initial breach disclosure. In an update dated 13 November 2026, Notion said it had reported exposed third-party API keys found in affected user messages to the vendors that issued them. The company also said it revoked exposed Notion API keys found during its investigation on 11 November.
The incident is significant for SaaS companies because AI features often rely on multiple external providers. A customer may interact with one application without knowing that their requests, responses or other data are processed by another company's infrastructure.
How the Fireworks incident happened
According to Fireworks AI's security bulletin, a third-party cloud provider hosting its research and development machines experienced a security incident on 23 October 2026. On 24 October, an unauthorized party used that access to reach Fireworks development machines and run an unauthorized workload. Fireworks said it detected the activity that day and began remediation.
The company's 9 November update identified several categories of potentially affected information:
- Resource metadata, including resource identifiers and descriptions.
- Billing information, such as billing and shipping addresses, invoice history, account balances and credit usage.
- A subset of incomplete model weights and datasets for certain customers.
Fireworks said it does not store payment card numbers and that those numbers were not affected. It also said it had not detected an impact to its Inference, Training SDK and Nexus services, which it describes as zero-data-retention by default. The investigation was still ongoing when that bulletin was published.
Why Notion customers were affected
Notion used Fireworks for classification jobs supporting its Insights feature, which helps administrators understand how members use AI. It also used the provider for inference involving open-weight models.
Notion's timeline shows how the scope of the incident became clearer over several days:
| Date | Reported development |
|---|---|
| 2 November | Fireworks notified Notion that a small dataset had been accessed. |
| 3 November | Notion disconnected its systems from Fireworks and deleted its data there. |
| 6 November | Fireworks notified Notion that customer data may have been accessed. |
| 8 November | Fireworks confirmed the final dataset relevant to Notion. |
| 9 November | Notion contacted affected customers. |
| 11 November | Notion revoked exposed Notion API keys found in affected messages. |
| 12 November | Notion reported exposed third-party API keys to the vendors that issued them. |
| 13 November | Notion published the latest update describing the key-related actions. |
Notion says the incident affected fewer than 1% of its workspaces. The exposed material included chat transcripts associated with the Insights feature. For the vast majority of affected workspaces, the transcripts were not labelled with the identity of the user or the workspace, and they did not include tool-call results. However, the content of a conversation can itself contain identifying or sensitive information.
Notion has said that no other Notion features were affected. It also disabled Insights during the investigation and stopped serving open-weight models through Fireworks.
What the API-key update means
The 13 November update deserves particular attention from engineering and security teams.
Notion said that exposed third-party API keys found in affected user messages were reported to the vendors that issued them. It separately said that exposed Notion API keys discovered during the investigation had been revoked on 11 November. These are distinct actions: the update does not establish that every third-party key was issued by Notion or that every affected user had a credential exposed.
For customers who received an incident notification, Notion provides a security-incident review process through the workspace interface. The company recommends treating affected content as exposed and rotating passwords, API keys or other credentials that may have been included in the relevant AI conversations.
Lessons for SaaS teams using external AI providers
This incident highlights several controls that should be considered part of an AI integration rather than optional security extras.
First, minimize the information sent to model providers. Applications should avoid transmitting entire conversation histories or unrelated workspace data when a smaller payload can accomplish the task.
Second, distinguish inference from data-processing workflows. Real-time model requests, batch classification, evaluation datasets and model-training jobs may have different storage and retention characteristics. A zero-data-retention setting for one service should not be assumed to cover every product or workflow.
Third, establish credential-handling controls. Applications should detect API keys and secrets before sending content to external AI services. Where appropriate, teams should redact secrets, use short-lived credentials and maintain a documented rotation process.
Finally, maintain a current inventory of AI subprocessors. Teams should know which vendors receive customer data, what they retain, where they process it, and how they notify customers when an incident occurs.
What to watch next
Fireworks said its investigation and security hardening were continuing. Notion also described its investigation as ongoing and said it would contact customers if additional workspaces were identified as affected. Further updates may clarify the number of impacted records, the scope of credential exposure and whether additional vendor notifications are required.
For SaaS businesses, the central lesson is straightforward: security commitments made by a primary application are only as dependable as the data-handling practices of the providers supporting it.
Why this is trending
- Hacker News: The Fireworks–Notion disclosure received 165 points and 66 comments in the retrieved discussion. Comments focused on third-party vendor risk, the meaning of zero data retention, and whether conversation content can remain identifying even without account metadata.
- TechCrunch: Published a report on 12 November detailing the breach and the exposure of Notion AI chat transcripts.
- Notion's own security updates: The 13 November update added a material development about exposed API keys, making this more than a repeat of the original breach report.
- X: Search access returned 403; no reliable discussion sample could be verified.
- Reddit: Direct search access returned 403; no reliable topic-specific discussion sample could be verified.
- daily.dev: The public site was accessible, but its topical feed required authenticated access. No verified topic-specific signal was retrieved.
The strongest verified signals were the Notion disclosure, TechCrunch reporting and substantive Hacker News discussion.
Sources
- Notion — Fireworks AI incident security bulletin — Updated 13 November 2026. Primary source for the affected workspaces, incident timeline and latest API-key actions.
- Notion — Notice about the Fireworks AI incident — Current help-centre guidance. Explains affected-customer review steps and recommended credential rotation.
- Fireworks AI — Update on October 2026 security incident — Updated 9 November 2026. Primary source for the cloud-provider compromise, affected data categories and Fireworks' remediation claims.
- TechCrunch — Cyberattack on Fireworks AI allowed hackers to access Notion customers’ chat transcripts — 12 November 2026. Independent reporting on the incident and its customer impact.
- Hacker News — Fireworks AI security breach impacts Notion and their customers — November 2026. Community discussion and reactions.
Further reading
- Notion — Fireworks AI incident security bulletin — Updated 13 November 2026. Primary source for the affected workspaces, incident timeline and latest API-key actions.
- Notion — Notice about the Fireworks AI incident — Current help-centre guidance. Explains affected-customer review steps and recommended credential rotation.
- Fireworks AI — Update on October 2026 security incident — Updated 9 November 2026. Primary source for the cloud-provider compromise, affected data categories and Fireworks' remediation claims.
- TechCrunch — Cyberattack on Fireworks AI allowed hackers to access Notion customers’ chat transcripts — 12 November 2026. Independent reporting on the incident and its customer impact.
- Hacker News — Fireworks AI security breach impacts Notion and their customers — November 2026. Community discussion and reactions.
Building a product in this space?
MORE FROM THE NEWSROOM
Related news
CybersecurityMalicious Optimizely Scripts Reached 178 Organizations, Exposing Website Visitors to Data Theft
A malicious JavaScript campaign distributed through Optimizely Web reached visitors across 178 organizations, according to security researcher Scott Helme. The payload included cryptocurrency address substitution, payment-card skimming, cookie theft and iOS exploit delivery mechanisms, highlighting the security risks associated with trusted third-party website scripts.
CybersecurityUber Driver Information Exposed After Fireworks AI Vendor Breach, State Filings Reveal
Uber has confirmed that personal information belonging to drivers was accessed through compromised credentials at a third-party technology provider identified as Fireworks AI. Public breach notices and independent reporting identify at least 7,870 affected individuals across Massachusetts, South Carolina and Rhode Island, with additional state notifications still being assessed.
CybersecuritySignDrop Windows Zero-Day Exposes Local SYSTEM Privilege-Escalation Path; Microsoft Investigates
Security researcher Abdelhamid Naceri has disclosed SignDrop, a Windows privilege-escalation vulnerability involving the Common Log File System and Microsoft Defender. A public proof of concept demonstrates a path from an existing local user account to SYSTEM privileges, although Microsoft says Defender detects and blocks the associated exploit behavior and that it has not observed exploitation in the wild.
