🔍 Read the full analysis: Notion AI Transcript Exposure: The Fireworks Breach And Retention Promise 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 Notion AI chat transcripts used for its Insights feature, affecting fewer than 1% of workspaces. The company says Fireworks had made a contractual zero-data-retention commitment, but Notion later learned data was retained; it has not identified which system stored it or for how long.
Notion says a dataset accessed by attackers in the Fireworks AI breach included chat transcripts from Notion AI’s Insights feature, and that the data had been retained by Fireworks despite a contractual zero-data-retention commitment. Notion estimates 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 Fireworks processed data for two purposes: classification jobs for Insights, which gives workspace administrators information about AI use, and inference for open-weight models in Notion’s product. The exposed transcripts came from the Insights work. Notion says the dataset Fireworks provided on November 8 included chat transcripts used for that feature.
For most affected workspaces, Notion says the transcripts were “unattributable”: 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 including model inputs and outputs written in the chat, which may contain sensitive details entered by users. The company advises affected customers to review the data as if it were exposed.
Notion says AI Meeting Notes, including audio, transcripts and summaries, were not affected unless that material was used as an input in a Notion AI chat transcript. It says no other Notion features were affected. The company also says it revoked exposed Notion API keys found in affected messages; users should review incident guidance and rotate passwords, API keys or other credentials they pasted into Notion AI.
Notion AI Transcript Exposure The Fireworks breach and retention promise
A dataset accessed in the Fireworks AI breach included some Notion AI Insights chat transcripts. Notion says data was retained despite a contractual zero-data-retention commitment.
What was exposed
The dataset contained transcripts sent for Insights analysis.
Chat inputs and outputs
Notion says transcripts include text written into chats, which may contain sensitive details.
Most other features
Notion says no other features were affected. Meeting Notes were excluded unless used as input in an AI chat transcript.
The retention gap
Live model requests and data stored for product features followed different retention terms.
- Fireworks agreed to zero data retention for Notion’s platform use.
- Notion says live requests were covered by separate terms from stored Insights data.
- Data from Fireworks endpoints used for Insights had been retained.
- The system and retention period have not been identified.
Incident timeline
What affected customers can do
Notion advises notified customers to treat affected material as exposed.
- Review incident guidance and the material entered into Notion AI.
- Rotate passwords, API keys, or other credentials pasted into chats.
- Workspace owners can decide whether further internal security steps are needed.
What remains unclear
What the Retention Gap Means
The incident puts a practical limit on how customers may understand zero data retention. Notion says that commitment covers live requests sent to model vendors to generate responses, but does not cover data stored to provide features such as chat history and Insights. For Insights, Notion says it sent stored requests and responses to Fireworks for analysis, under a separate contractual commitment that data would not be retained.
Notion now says it learned that data from the Fireworks endpoints it used was retained. The distinction matters to organizations that rely on vendor promises when deciding what employees can enter into AI tools. A contractual commitment and the system’s actual handling of data did not align in this case, according to Notion. The company says it is auditing critical vendors and their data-minimization practices; the public information does not establish whether the retention was intentional or how it occurred.
For affected users, the exposure risk depends on what was written in chats, not only whether a transcript was linked to a person or workspace. Notion specifically recommends reviewing affected data because inputs and outputs may contain sensitive material. Its estimate of fewer than 1% of workspaces is not a count of individuals, and the company has not published a total number of affected users.
Popular right now
Shark Pet Cordless Stick Vacuum, Up to 40-Minute Runtime, Lightweight, IX141
Samsung 43-Inch M70H Series 4K Mini LED Smart TV (2026)
Roborock RockNeo Q110H Wire-Free Robot Lawn Mower
As an affiliate, we earn on qualifying purchases.
How Notion Used Fireworks
Notion’s November advisory updates a sequence of responses to Fireworks’ breach. According to Notion’s published timeline, Fireworks first notified the company on November 2 that a small set of data had been accessed by malicious actors. Notion initially understood the material to be synthetic testing data. On November 3, it says, it disconnected its systems from Fireworks, deleted data in the Fireworks system and halted training runs.
Notion’s notice changed after it received additional information and reviewed the dataset. Its updates on November 10 and 11 added the transcript exposure and retention details. Fireworks’ own bulletin, as described in the source material, remained marked “Last updated November 9.” Fireworks said it had not detected an impact to Inference, Training SDK and Nexus, which it described as zero-retention by default, while listing datasets among data accessed for certain customers. The published statements do not identify which Fireworks system held Notion’s data.
Notion says its zero-retention terms for live model requests are distinct from data stored for features that require it. The exposure discussed here concerns stored requests and responses used for Insights analysis, rather than a blanket claim that all Notion AI conversations or features were affected.
““Unfortunately, during our investigation we learned that data was retained from the Fireworks endpoints that we used.””
— Notion, in its updated security advisory
What the Notices Do Not Explain
Notion has not said which Fireworks system retained the data, how long it remained there, or how many individual people are represented in the affected workspaces. Its estimate of fewer than 1% of workspaces does not provide a precise user count. The company also has not explained why the retention occurred despite the contractual commitment.
The public notices do not establish whether other Fireworks customers experienced a similar retention issue. Fireworks’ statement about zero-retention-by-default services and its listing of datasets accessed do not resolve which system handled Notion’s stored Insights data. The companies’ published accounts leave that point open, so it cannot be determined from the available material whether the statements refer to different systems or different data-handling paths.
Notion says that if neither a user nor their workspace owner has received a notice, it has not identified that workspace as affected. That is the company’s current identification status; the available information does not provide an independent accounting of the full dataset or confirm whether further affected workspaces could be identified.
Vendor Audits and Customer Reviews
Notion says it is auditing critical vendors, including inference providers and data processors, to check whether they follow data-minimization practices. It has not announced a completion date or published findings from that review. Further updates from Notion or Fireworks may clarify which endpoint retained the data, the retention period and whether other customers were affected.
Customers who have been notified should follow Notion’s Security incident review guidance, treat affected material as exposed and rotate credentials pasted into Notion AI during the incident. Workspace owners can review what was entered into the service and decide whether additional internal security steps are needed. Notion’s advice is directed to affected workspaces; the company says it has not identified a workspace as affected if neither the user nor its owner has received notice.
Key Questions
What Notion AI data was exposed?
Notion says the dataset accessed by attackers included chat transcripts used for the Insights feature. It defines transcripts as model inputs and outputs written in the chat. The company says AI Meeting Notes were not affected unless their content was used as input in a Notion AI chat transcript.
How many Notion workspaces were affected?
Notion estimates that fewer than 1% of workspaces were affected. It has not published a precise number of workspaces or individuals.
Did the transcripts identify their users?
Notion says transcripts for the vast majority of affected workspaces were unattributable, meaning they were not labeled with the person who wrote them or the workspace they came from. The text could still contain sensitive information entered into a chat.
What should notified users do?
Notion advises affected customers to review the data as if it were exposed. Users should follow the company’s Security incident review guidance and rotate passwords, API keys or other credentials they pasted into Notion AI during the incident.
Why was data retained if Fireworks promised zero retention?
Notion says Fireworks made a contractual zero-retention commitment, then says it learned data from the endpoints it used had been retained. The published notices do not explain which system stored the data, why it was retained or for how long.
Source: ThorstenMeyerAI.com
