AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Notion’s AI Retention Promise Meets A Fireworks Breach Disclosure on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Notion says data accessed in the Fireworks AI breach included Notion AI chat transcripts used for its Insights feature, affecting fewer than 1% of workspaces. The company says Fireworks had promised zero data retention but retained data from endpoints Notion used; details about where and for how long remain unknown.

Notion says data accessed in the Fireworks AI breach included Notion AI chat transcripts used for its Insights feature, and that Fireworks retained data despite a contractual zero-data-retention commitment. Notion estimates that fewer than 1% of workspaces were affected and says it revoked exposed Notion API keys found in affected user messages on November 11.

In its updated security advisory and Help Center notice, Notion said a dataset Fireworks provided on November 8 included chat transcripts associated with Insights, a feature that helps workspace administrators understand how teams use AI. Notion said the transcripts were unattributable for most affected workspaces: they were not labeled with the author or workspace and did not include tool-call results. That does not mean the text was necessarily harmless. Notion defines a transcript as any input to or output from a model written in the chat, which could include sensitive details users entered.

Notion said fewer than 1% of workspaces were affected. It advised affected customers to treat the data as exposed, review it through its Security incident review guidance, and rotate passwords, API keys or other credentials pasted into Notion AI during the relevant period. On November 11, the company said it revoked exposed Notion API keys found in affected user messages. Notion says it has not identified a workspace as affected if neither the workspace owner nor the user has received a notice from the company.

Notion said AI Meeting Notes—including audio, transcripts and summaries—were not affected unless they were used as input in a Notion AI chat transcript. It also said no other Notion features were affected. Those statements are the company’s account of the incident; the public notice does not provide a count of affected people or a detailed inventory of exposed transcript contents.

At a glance
updateWhen: Notion updated its notice on November 1…
The developmentNotion expanded its security notice on November 10 and 11 to say Fireworks breach data included some Notion AI chat transcripts and that data was retained despite a contractual zero-retention commitment.
AI Security · Briefing

Notion’s AI retention promise meets a Fireworks breach

Notion says Fireworks retained data despite a zero-retention commitment, and breach data included some Notion AI chat transcripts used for Insights.

<1%
of workspaces affected, per Notion
November 8
Fireworks provided the dataset
November 11
Notion said it revoked exposed API keys

What the breach data included

Notion says the dataset included chat transcripts associated with Insights.

Transcript data

Chat inputs and outputs

Transcripts can contain details users entered, even when records lack author or workspace labels.

Not included

Most other features

Notion says AI Meeting Notes and other features were not affected, unless meeting content was used in a chat transcript.

Why the retention promise matters

Notion says Fireworks made a contractual zero-retention commitment, but data from the endpoints Notion used was retained.

Retention terms
  • Notion says vendor ZDR covers live requests to generate AI responses.
  • Stored requests and responses for chat history and Insights fall outside that commitment.
What remains unclear
  • Which Fireworks system retained the data and for how long.
  • How many people or transcripts were affected, or what each transcript contained.

What affected workspaces should do

Notion advises customers who receive a notice to treat relevant data as exposed.

1
Review the data

Use Notion’s Security incident review guidance to examine affected chats.

2
Rotate credentials

Change passwords, API keys, or other credentials pasted into Notion AI.

3
Check for notices

Notion says it has not identified a workspace as affected if neither its owner nor user received a notice.

Open questionsNotion has not disclosed a precise count, date range, or detailed inventory of exposed transcript contents.
The takeaway
A zero-retention promise did not prevent retention here; affected customers should review exposed chats and rotate any credentials they contained.

What Transcript Exposure Means for Workspaces

The disclosure matters because chat transcripts can contain user-entered information, even when they lack names or workspace labels. Notion recommends that affected customers review the specific data, since prompts or model responses may include sensitive material. The company’s description of transcripts as unattributable may reduce the ease of linking some records to individuals, but it does not establish that all content was non-sensitive or impossible to identify.

The incident also puts attention on a distinction between a vendor’s zero-retention commitment and the data handling required to provide a product feature. Notion says its ZDR commitments cover live requests to model vendors, while stored requests and responses used for chat history and Insights fall outside that scope. In this case, Notion says Fireworks separately made a contractual ZDR commitment for its platform, but data from the endpoints Notion used was retained. Customers relying on AI vendors’ retention terms may want to understand which data flows and services those terms actually cover.

For affected organizations, the practical concern is not just whether a transcript was accessed, but what it contained and whether it included credentials or confidential business information. Notion’s instruction to treat the data as exposed is a precaution; the notice does not say that every exposed transcript contained sensitive information or that credentials were used by attackers.

Products Worth Considering

Amazon Kindle Scribe Colorsoft 64GB with 11-Inch Color Display and Premium Pen, Graphite

Amazon Kindle Scribe Colorsoft 64GB with 11-Inch Color Display and Premium Pen, Graphite

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation

As an affiliate, we earn on qualifying purchases.

How Notion Used Fireworks

Notion said it used Fireworks for two purposes: classification jobs for Insights and as a general inference provider for open-weight models in its product. The newly disclosed transcripts came from the Insights-related dataset, according to Notion. The company says it disconnected its systems from Fireworks on November 3, deleted data in the Fireworks system and stopped communications between the services.

Notion’s published incident timeline says Fireworks first notified it on November 2 that a small set of data had been accessed by malicious actors; Notion initially understood the data to be synthetic testing data. The source material says Fireworks’ own bulletin remained marked as last updated November 9, while Notion updated its notice on November 10 and 11. BleepingComputer reported the transcript disclosure on November 11.

Notion says its ZDR commitments for model vendors apply to API requests sent to generate Notion AI responses. It says uses that require retained information, including chat history and Insights, are outside that particular commitment. Separately, Notion says Fireworks made a contractual commitment to zero data retention for Notion’s use of its platform.

““Unfortunately, during our investigation we learned that data was retained from the Fireworks endpoints that we used.””

— Notion, in its updated security notice

Open Questions About Retained Data

The public notices do not identify which Fireworks system retained Notion’s data, how long it was retained, or how many individual users were affected. Notion has not published a transcript-by-transcript account of what information was present, and the available material does not establish whether attackers viewed or used credentials contained in messages.

There is also an unresolved difference in emphasis between the companies’ statements. Fireworks’ November 9 update says it detected no impact to Inference, Training SDK and Nexus, which it describes as zero-retention by default, while listing datasets among data accessed for certain customers. Notion says data was retained from the Fireworks endpoints it used. The published information does not explain which systems or data paths these statements refer to, so it is not possible to reconcile them or determine whether other Fireworks customers faced the same issue.

Notion’s estimate applies to workspaces, not a published count of people or transcripts. The company has not disclosed a precise total, the date range covered by the exposed material, or whether any additional data categories will be identified as its investigation continues.

Vendor Audit and Customer Reviews

Notion says it is auditing critical vendors, including inference providers and data processors, to check their data-minimization practices and adherence to zero-retention policies. It has not announced a completion date or described the audit’s scope beyond that statement. The company’s notice and help page may be updated as it learns more.

Workspace owners who receive a notice from Notion should follow its incident-review guidance and examine whether affected chats included passwords, API keys or other credentials. Notion’s stated advice is to rotate credentials pasted into Notion AI during the incident and treat the relevant data as exposed. Customers who have not been notified should not infer that their workspace was identified as affected, according to Notion, though the company has not published a full dataset or independent verification of its affected-workspace estimate.

Further clarity depends on additional information from Notion and Fireworks about the retained data, the systems involved and the investigation’s findings. Neither company’s available notice settles whether other customers’ data was retained in a similar way.

Key Questions

What Notion data was involved?

Notion says the dataset included Notion AI chat transcripts used for Insights. It says transcripts for most affected workspaces were not labeled with a user or workspace and did not contain tool-call results.

How many workspaces does Notion say were affected?

Notion estimates fewer than 1% of workspaces were affected. It has not provided a precise workspace count or a number of individual users.

What should affected workspace owners do?

Notion advises affected customers to treat the data as exposed, review it using the company’s Security incident review guidance, and rotate passwords, API keys or other credentials pasted into Notion AI during the incident.

Were AI Meeting Notes affected?

Notion says AI Meeting Notes—including audio, transcripts and summaries—were not affected unless they were used as input in a Notion AI chat transcript. It says no other Notion features were affected.

What is still unknown about Fireworks’ data retention?

The public notices do not specify which Fireworks system retained the data, how long it was kept, or whether other Fireworks customers experienced the same problem. The companies’ published descriptions do not fully explain how Notion’s retention disclosure relates to Fireworks’ statement about zero-retention services.

Source: ThorstenMeyerAI.com

You May Also Like

Western Digital Surges In Global Coverage

Western Digital experiences a surge in worldwide media mentions, with 33 reports in a recent window, indicating increased global attention.

Inside The Ex-ByteDance Intern’s $200M AI Lab And Its Fei-Fei Li Rivalry

A Moneyweb headline describes a former ByteDance intern’s $200 million AI lab as taking on Fei-Fei Li, but key details are unavailable.

Should You Use Mistral Forge? A Buyer’s Decision Guide

An analysis of Mistral Forge’s suitability for enterprises, highlighting when it fits and when alternatives are better, based on current expert insights.

Glasspane: When Transparency Itself Becomes the Product

Glasspane introduces role-aware dashboards and AI-driven insights, redefining infrastructure transparency for enterprises and MSPs.