🔍 Read the full analysis: Fireworks Breach Puts Notion AI Conversations In The Privacy Spotlight on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Notion says a dataset accessed in the Fireworks AI breach included chat transcripts from its AI Insights feature, affecting fewer than 1% of workspaces. On November 11, Notion revoked exposed API keys found in affected messages; it advises affected users to treat the data as exposed and review it for credentials and other sensitive content.
Notion says chat transcripts from its AI Insights feature were included in a dataset accessed by attackers in the Fireworks AI breach, and that Fireworks retained data despite a contractual commitment to zero data retention. Notion estimates fewer than 1% of workspaces were affected and says it revoked exposed Notion API keys found in affected user messages on November 11.
The transcripts were part of data sent to Fireworks for Insights, a Notion feature that gives workspace administrators information about how teams use AI. Notion says the dataset Fireworks provided on November 8 included Notion AI chat transcripts used for that feature. It describes most affected transcripts as unattributable: they were not labeled with the author or workspace, and did not include tool-call results.
That description does not mean the transcript contents were necessarily harmless or anonymous in practice. Notion defines a transcript as input to or output from the model that was written to the chat. Users may have typed customer names, contract information, passwords, API keys or other sensitive material into those chats. Notion advises affected organizations to review the data because individual inputs and outputs may contain sensitive information.
Notion says it disconnected its systems from Fireworks on November 3, deleted data in the Fireworks system and stopped communications with the provider. On November 11, it says it revoked exposed Notion API keys identified in affected user messages. The company recommends that affected users review the material under its Security incident review guidance and rotate credentials they entered into Notion AI during the incident.
Fireworks breach puts Notion AI chats in focus
Notion says attackers accessed a dataset that included AI Insights chat transcripts. Affected organizations should treat the data as exposed and review chats for sensitive content.
What was exposed
The accessed dataset included transcripts sent to Fireworks for Notion’s AI Insights feature.
Chat input and output
A transcript can contain anything written into a prompt or returned in a chat, including sensitive details.
Labels were absent
Most affected transcripts lacked author and workspace labels and did not include tool-call results.
What Notion says happened
Notion says Fireworks retained data despite a contractual zero-data-retention commitment.
What remains unclear
The available notices leave key details unresolved.
- Which Fireworks system retained the data, and for how long.
- Why retention occurred despite the stated commitment.
- Whether attackers viewed specific chats or other customers were affected.
- What sensitive information was present in individual transcripts.
What affected organizations can do
Notion says identified affected workspaces should have received notice.
Use Notion’s Security incident review guidance to check affected material for sensitive content.
Rotate passwords, API keys, and other credentials entered into Notion AI during the incident.
Why Retained AI Chats Matter
The disclosure matters because it concerns the gap between a vendor’s data-handling commitment and what happened to data processed for a customer feature. Notion says Fireworks had contractually committed to zero data retention for Notion’s use of its platform, but the investigation found that data from the endpoints Notion used had been retained. The companies have not publicly explained how or where that retention occurred.
For customers, the risk is not limited to whether a chat can be linked to a named user or workspace. A transcript can contain its own identifying or confidential details, including material pasted into a prompt. Notion’s estimate that fewer than 1% of workspaces were affected narrows the stated scope, but does not establish what sensitive information was present in any particular workspace.
The incident also makes the scope of “zero data retention” relevant to buyers of AI products. Notion says its ZDR commitments apply to live requests sent to model providers to generate responses, but not to data stored for features such as chat history and Insights. That distinction, and whether vendors follow their commitments in practice, can affect how organizations assess AI tools and handle sensitive information.
Popular right now

Ninja Air Fryer XL, 6.5 Qt, 6-in-1, MaxCrisp, Dark Grey (AF181)

Vitamix 5200 Blender with 64 oz Container, Black

Dyson V8 Cordless Vacuum
As an affiliate, we earn on qualifying purchases.
How Insights Used Fireworks
Notion says it used Fireworks in two ways: for classification jobs supporting the Insights feature, and as a general inference provider for open-weight models within Notion. The company says the exposed dataset came from the Insights work. It also says the stored requests and responses from Notion AI use were sent to Fireworks for analysis, rather than being limited to live requests for model responses.
Notion’s published timeline says Fireworks first notified it on the evening of 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 company says it disconnected from Fireworks and removed its data from the Fireworks system on November 3. Notion updated its advisory and help materials on November 10 and 11 to describe the transcript exposure and response.
Fireworks’ own bulletin, according to the source material, still showed a last-updated date of November 9. Fireworks said it had not detected an impact to Inference, Training SDK and Nexus, which it described as zero-data-retention by default, while listing datasets among data accessed for certain customers. That statement does not identify which system retained Notion’s data.
“Unfortunately, during our investigation we learned that data was retained from the Fireworks endpoints that we used.”
— Notion
Open Questions on Fireworks Retention
The available notices do not identify which Fireworks system retained Notion’s data, how long the data was kept, or why retention occurred despite the stated contractual commitment. The materials also do not give a precise number of affected people or workspaces beyond Notion’s estimate that fewer than 1% of workspaces were affected.
It remains unclear whether other Fireworks customers experienced the same retention issue. Fireworks’ statement about no detected impact to certain services does not resolve the question of which system held the Notion data, and the published information does not reconcile that statement with Notion’s account. The source material also does not establish whether attackers viewed or used specific chat contents, or what sensitive information was present in each affected transcript.
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. Those limits are based on the company’s current notice; further technical findings have not been provided in the supplied material.
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 zero-retention policies. It has not provided a public timetable for completing that audit. Further updates from Notion or Fireworks could clarify the system involved, the length of retention and whether the issue affected other customers.
Notion says workspaces it has identified as affected should have received notice. Organizations that receive one should review the relevant chats under Notion’s Security incident review guidance, treat the data as exposed and rotate any passwords, API keys or other credentials entered during the incident. Notion says it has already revoked exposed Notion API keys found in affected user messages, but users should still review other credentials and sensitive material in their own chats.
Key Questions
What Notion AI data was included in the accessed dataset?
Notion says the dataset included chat transcripts used for its Insights feature. It defines a transcript as model input or output written to the chat. Most affected transcripts were described as unattributable, meaning they lacked labels identifying the author or workspace and did not contain tool-call results.
How many Notion workspaces were affected?
Notion says fewer than 1% of workspaces were affected. The supplied information does not give a precise count of workspaces or people. Notion says that if neither a user nor their workspace owner has heard from the company, it has not identified that workspace as affected.
What should an affected workspace do?
Notion advises affected users to treat the data as exposed, review it under the company’s Security incident review guidance and rotate passwords, API keys or other credentials pasted into Notion AI during the incident. Notion says it revoked exposed Notion API keys found in affected messages on November 11.
Did the breach affect Notion AI Meeting Notes?
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. The company says no other Notion features were affected.
Why was data retained despite a zero-retention commitment?
That remains unclear. Notion says Fireworks had made a contractual zero-retention commitment, but its investigation found that data from the endpoints it used had been retained. The published information does not identify the Fireworks system involved or explain why the data was kept.
Source: ThorstenMeyerAI.com
