🔍 Read the full analysis: Did A Fireworks Security Breach Expose Notion AI Chat Transcripts? 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 data accessed in a Fireworks AI breach included Notion AI chat transcripts used for its Insights feature, affecting fewer than 1% of Notion workspaces. The company says it had a contractual zero-retention commitment from Fireworks but later learned that data from the endpoints it used had been retained. Notion revoked exposed API keys on November 11 and advises affected users to review the data and rotate credentials.
Notion says chat transcripts from its AI service were included in a dataset accessed by attackers in a breach at AI provider Fireworks, and that on November 11 it revoked exposed Notion API keys found in affected user messages. Notion says fewer than 1% of its workspaces were affected; it has not published the number of people involved or identified the Fireworks system that retained the data.
In its updated security advisory and Help Center notice, Notion said the exposed dataset included Notion AI chat transcripts used for its Insights feature. Insights gives workspace administrators information about how teams use AI. Notion said it received the dataset from Fireworks on November 8 and that, for most affected workspaces, the transcripts were not labeled with the author or workspace and did not contain tool-call results.
Notion defines a transcript as any input to or output from the model written in the chat. That means messages may contain information users typed or received, including potentially sensitive material. The company advises affected customers to review the data and treat it as if it were exposed. It recommends rotating passwords, API keys, or other credentials that were pasted into Notion AI during the incident.
Notion said the incident affected fewer than 1% of workspaces. It also said AI Meeting Notes, including audio, transcripts, and summaries, were not affected unless they had been used as input in a Notion AI chat transcript. The company said no other Notion features were affected. On November 11, it revoked exposed Notion API keys found in affected messages.
Notion AI Chats in Breach What Users Need to Know
Notion says transcripts used for its Insights feature were included in data accessed during a Fireworks breach. Affected users should review the data and rotate credentials shared in chats.
What the dataset contained
Notion says the transcripts came from Insights classification work.
Inputs and outputs
Transcripts may include text users entered or received, including sensitive material.
Often unlabeled
Most transcripts were not labeled with an author or workspace and lacked tool-call results.
Limited feature impact
Notion says no other features were affected; AI Meeting Notes were unaffected unless used as chat input.
Why retention is in question
Notion says Fireworks had committed to zero retention, but data from the endpoints Notion used was later found to have been retained.
- Fireworks had a contractual zero-retention commitment.
- Notion learned data from its endpoints had been retained.
- Which Fireworks system retained the data.
- How long it was kept or why.
What affected users can do
Notion advises affected customers to treat the data as exposed.
Notion says affected workspace owners or users will be notified.
Look for sensitive information in the affected transcripts.
Change passwords, API keys, or other credentials pasted into Notion AI during the incident.
Why Retained AI Data Matters
The disclosure matters because Notion says Fireworks had made a contractual commitment to zero data retention for Notion’s use of its platform. Notion later said it learned that data had nevertheless been retained from the Fireworks endpoints it used. The incident therefore concerns not only unauthorized access, but also a gap between the stated retention commitment and what happened to data sent for processing.
For customers, the main practical risk is the possible content of chats. Notion says transcripts may contain any text entered into the chat or generated in response, and recommends a case-by-case review for sensitive information. Its statement that most transcripts were unattributable may limit direct links to an individual or workspace, but it does not establish that their contents were harmless or impossible to identify from context.
The episode also raises questions for companies relying on AI vendors and their retention terms. Notion says it is auditing critical vendors and data processors to verify data-minimization practices, including zero-retention policies. The published information does not establish whether the issue resulted from a particular system, configuration, or other cause.
Products Worth Considering
Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black
Logitech MX Mechanical Wireless Illuminated Keyboard, Clicky Switches
Samsung 32 Odyssey G5 (G51F) QHD 180Hz Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How Notion Used Fireworks
Notion said it used Fireworks in two ways: to run classification jobs for Insights and as an inference provider for open-weight models within its product. The transcripts described in the advisory came from the Insights work. Notion said it sent stored requests and responses from Notion AI use to Fireworks for analysis.
Notion’s notice distinguishes those stored records from its zero-retention commitments for live model requests. It says those commitments apply to API requests sent to third-party vendors to generate Notion AI responses, but are not intended to cover features that require retaining data, such as chat history and Insights. Notion separately said Fireworks had promised zero retention for Notion’s use of its platform, then reported that data from the endpoints it used had been retained.
In its November 9 update, Fireworks said it had not detected an impact to Inference, Training SDK, and Nexus, which it described as zero-retention by default. It also listed datasets among the types of data accessed for certain customers. Notion’s updated account adds the transcript disclosure; the two companies’ public statements do not identify which Fireworks system retained Notion’s data.
“Unfortunately, during our investigation we learned that data was retained from the Fireworks endpoints that we used.”
— Notion, in its updated security notice
What the Public Notices Do Not Explain
The affected count is not public beyond Notion’s estimate that fewer than 1% of workspaces were affected. The company has not disclosed how many users or transcripts that represents, which Fireworks system retained the data, how long it was kept, or why it was retained despite the stated commitment.
Notion’s and Fireworks’ descriptions also leave an unresolved question about how the retained data relates to the services Fireworks says were not affected. Neither company’s published account identifies the system or process involved, so the available information does not support a more specific explanation. It is also unclear whether other Fireworks customers experienced a similar retention issue.
Notion says affected workspace owners or users will be notified; it says a workspace that has received no notice has not been identified as affected. The company has not said when its vendor audit will be completed or whether it will publish further findings.
Notices, Credential Reviews and Vendor Audit
People who receive a notice from Notion should follow the company’s incident-review guidance, examine affected data, and rotate any credentials pasted into Notion AI during the incident. Workspace owners can use the notice to determine whether Notion has identified their workspace as affected; the company says it has not identified workspaces that have not received a notification.
Notion says it is auditing critical vendors, including inference providers and data processors, to check data-minimization practices and zero-retention policies. The next useful public updates would clarify the Fireworks system involved, the period for which data was retained, and whether the review changes the stated number or scope of affected workspaces. Fireworks’ cited bulletin was last updated November 9, while Notion’s advisory was updated November 10 and 11.
Key Questions
Did the Fireworks breach expose Notion AI chat transcripts?
Notion says it did. Its updated advisory says the dataset accessed in the Fireworks incident included transcripts used for Notion’s Insights feature.
How many Notion workspaces were affected?
Notion estimates fewer than 1% of workspaces were affected. It has not published a precise count of workspaces, users, or transcripts.
Were all Notion AI conversations exposed?
Notion has not said that all conversations were exposed. It identifies transcripts associated with the Insights feature and says the affected workspaces were fewer than 1% of the total.
What should affected users do?
Notion advises affected customers to review the data as if it were exposed and rotate passwords, API keys, or other credentials pasted into Notion AI during the incident.
What is still unknown about the incident?
The public notices do not specify which Fireworks system retained the data, how long it was retained, or how many people were affected. It is also unclear whether other Fireworks customers faced the same issue.
Source: ThorstenMeyerAI.com
