🔍 Read the full analysis: What Notion’s Fireworks Breach Report Means For AI Chat Privacy on ThorstenMeyerAI.com
Get business pricing on tech for your team
- 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 transcripts from Notion AI’s Insights feature, and that Fireworks retained data despite a contractual zero-retention commitment. Notion says fewer than 1% of workspaces are affected, but has not disclosed how long the data was retained or which Fireworks system held it.
Notion says data accessed in the Fireworks AI breach included chat transcripts from its AI Insights feature, and that Fireworks retained data despite a contractual zero-data-retention commitment. Notion estimates that fewer than 1% of its workspaces were affected and says it revoked exposed Notion API keys found in affected user messages on November 11.
Notion’s updated advisory says the dataset Fireworks provided to the company on November 8 contained Notion AI chat transcripts used for Insights, a feature that gives workspace administrators information about how teams use AI. Notion says transcripts from most affected workspaces were “unattributable”: they were not labeled with the writer’s identity or workspace, and did not include tool-call results. That does not mean the text itself was necessarily anonymous or harmless. Notion defines a transcript as model input or output written to the chat, which could include sensitive details entered by a user.
Notion says fewer than 1% of workspaces are affected. It says customers whose workspace has been identified will be contacted; if neither the user nor the workspace owner has received notice, Notion has not identified that workspace as affected. The company advises affected customers to treat the relevant data as exposed, review it under its Security incident review guidance, and rotate passwords, API keys or other credentials pasted into Notion AI during the incident. On November 11, Notion said it revoked exposed Notion API keys found in affected user messages.
Notion says the affected data came from the Insights analysis workflow, where stored AI requests and responses were sent to Fireworks for analysis. It says AI Meeting Notes—including audio, transcripts and summaries—were not affected unless their content was used as input in a Notion AI chat transcript. Notion also says no other Notion features were affected. Those statements describe the scope Notion has identified; the company has not published a count of affected users or the number of exposed transcripts.
What Notion’s Fireworks Breach Report Means For AI Chat Privacy
Notion says data from its AI Insights workflow was retained by Fireworks despite a zero-retention commitment. The exposed dataset included chat transcripts that could contain sensitive details.
What was exposed
Notion says the dataset involved AI Insights analysis, not general model inference.
AI chat transcripts
Transcripts may include model inputs or outputs written in a chat, including sensitive details.
Insights workflow
Stored requests and responses were sent to Fireworks for analysis.
Meeting Notes
Audio, transcripts and summaries were not affected unless used as input in a Notion AI chat.
Where the privacy gap appeared
Notion says live model requests and stored data for product features have different handling.
- Notion says model vendor zero-retention commitments apply to API requests used to generate responses.
- Stored requests and responses sent for analysis were retained from Fireworks endpoints, despite a contractual commitment.
What remains unclear
The public notices leave key details about the exposure unresolved.
| Question | Publicly disclosed |
|---|---|
| How long was data retained? | Not stated |
| Which Fireworks system held it? | Not identified |
| How many users or transcripts? | No count provided |
| Why did retention occur? | Not explained |
What customers can do
Notion says identified customers will be contacted.
- Affected customers should review relevant chats for sensitive content.
- Rotate passwords, API keys or other credentials entered in affected chats.
- Notion says exposed API keys found in affected messages were revoked on November 11.
What the Retention Failure Means
The incident matters because zero data retention is a condition some organizations rely on when deciding whether employees can use an AI service with business information. Notion says Fireworks made a contractual commitment to zero data retention for Notion’s use of its platform, yet its investigation found that data was retained from the Fireworks endpoints used. That is a reported gap between the commitment and what happened, though the public notices do not explain how or why it occurred.
There is also a distinction between a live AI request and the stored data used for a secondary feature. Notion says its zero-retention commitments for model vendors apply to API requests used to generate responses. It says features such as chat history and Insights require stored data and are outside that commitment. For Insights, Notion sent stored requests and responses to Fireworks for analysis. Readers evaluating AI privacy should therefore look not only at how a provider handles live prompts, but also at what data is retained or sent to vendors for analytics and other product functions.
For affected users, transcripts may contain information they chose to type into an AI conversation. Notion’s description that many records could not be linked to a particular person or workspace may reduce direct attribution, but it does not establish that the content lacked confidential material. The company itself recommends reviewing affected data for sensitive inputs and outputs.
Products Worth Considering
LG UltraGear 32-inch Curved QHD Gaming Monitor (2560×1440, 180Hz, 1ms, 1000R, HDR10)
Bose TV Speaker (2nd Gen) Compact Soundbar with Bluetooth and HDMI Connectivity, Black
Logitech MX Mechanical Wireless Illuminated Keyboard, Clicky Switches
As an affiliate, we earn on qualifying purchases.
How Notion Used Fireworks
Notion says it used Fireworks for two purposes: classification jobs for Insights and inference for open-weight models offered within Notion. According to Notion, the exposed dataset involved the Insights side, not the general inference function. Its latest advisory and help-page updates followed the breach investigation; the source account says Notion updated its notices on November 10 and 11.
Notion’s published timeline says Fireworks notified it on November 2 that a small set of data had been accessed by malicious actors. Notion initially identified the material as synthetic testing data. On November 3, it says it disconnected its systems from Fireworks, deleted data in the Fireworks system and halted communication and training runs. Fireworks’ own bulletin, as described in the source material, was last updated November 9 and says datasets were among the categories accessed for certain customers. The two companies’ descriptions do not identify which Fireworks system retained the Notion data.
Notion says it is auditing critical vendors, including inference providers and data processors, to check data-minimization practices and zero-retention policies. The available information concerns Notion’s use of Fireworks; it does not establish that other Fireworks customers had the same exposure.
“Unfortunately, during our investigation we learned that data was retained from the Fireworks endpoints that we used.”
— Notion, in its updated security FAQ
What the Notices Do Not Explain
The public notices do not say which Fireworks system retained the data, how long it was stored, or how many users and individual transcripts were involved. They also do not explain why data was retained despite the commitment Notion says it received, or whether the retention resulted from a particular feature, system setting or other cause.
Notion’s statement that fewer than 1% of workspaces were affected is a workspace-level estimate, not a disclosed count of people or conversations. The available information also does not establish whether other Fireworks customers experienced the same issue. Fireworks’ bulletin and Notion’s account differ in scope and detail, and the published material does not reconcile them; it would be premature to infer which system or process explains the difference.
Notion says most affected transcripts were unattributable, but has not supplied a breakdown of what kinds of sensitive information appeared in them. The notices also do not quantify how many exposed API keys were found or whether any were used by an unauthorized party.
Vendor Audit and Customer Reviews
Notion says it is auditing critical vendors and reviewing data-minimization practices, including zero-retention policies. The next useful update would clarify where the data was stored, for how long, and why, as well as whether the review found similar risks in other vendor workflows. No date for a further update or completion of the audit is given in the available source material.
Customers who have received an incident notice should follow Notion’s guidance to review relevant content and rotate credentials entered in affected chats. Notion says exposed API keys found in user messages were revoked on November 11. Workspace owners who have not received a notice should not assume their workspace has been identified as affected, but the company’s public estimate does not provide a detailed account of how notification decisions were made.
Key Questions
What Notion AI data was involved?
Notion says the dataset accessed in the Fireworks breach included chat transcripts used for the Insights feature. It defines transcripts as model input and output written to a chat.
How many Notion workspaces does the company say were affected?
Notion estimates that fewer than 1% of workspaces were affected. It has not published a count of users or transcripts.
Does Notion say AI Meeting Notes were exposed?
Notion says AI Meeting Notes, including audio, transcripts and summaries, were not affected unless their content was used as input in a Notion AI chat transcript.
What should an affected customer do?
Notion advises affected customers to treat the data as exposed, review it under its 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 user messages on November 11.
What remains unknown about Fireworks’ retention?
The public notices do not identify which Fireworks system retained the data, how long it was kept, why it was retained despite the commitment, or whether other Fireworks customers were affected.
Source: ThorstenMeyerAI.com
