Notion: Fireworks AI breach hit chat transcripts in under 1% of workspaces
A Fireworks AI breach exposed Notion AI chat transcripts in a small number of workspaces; affected customers should rotate any credentials typed into Notion AI.
Key points
- Notion said the incident at Fireworks AI is impacting less than 1% of Notion workspaces, and the company has contacted affected customers directly.
- The exposed data consisted of Notion AI chat transcripts used for the Insights feature, which has since been disabled.
- For most affected workspaces, Notion said the transcripts were unattributable, meaning they carried no labels linking them to specific users or organizations.
- Affected customers should rotate any passwords, API keys, or other credentials they may have pasted into Notion AI during the incident period.
Notion disclosed on November 12 that a security incident at Fireworks AI, one of its third-party vendors, exposed AI chat transcripts for a portion of its user base. The company said the breach touches less than 1% of Notion workspaces, and that it has already reached out to customers whose data was involved.
The incident did not originate inside Notion's own systems. Malicious actors broke into infrastructure at Fireworks AI, a platform Notion relied on for two purposes: running classification jobs that powered the Insights feature, and serving open-weight AI models inside the product. Notion said it is no longer routing model traffic through Fireworks, and it has disabled Insights while the investigation continues.
How the Fireworks AI vendor breach came to light
Events started on the evening of November 2, when Fireworks reached out to Notion to report that attackers had obtained a small set of data. At that point, Notion said the dataset turned out to include only synthetic testing data, with no actual customer information involved.
Moving quickly, Notion cut off its systems from Fireworks by the morning of November 3 and halted all data traffic between the two platforms. On November 4, it deleted all its data stored in the Fireworks system.
The picture changed on November 6, when Fireworks told Notion that malicious actors may have reached potential customer data. Fireworks confirmed the final scope of the exposed dataset on November 8. Notion then contacted impacted customers on November 9, three days before it published its public advisory.
BleepingComputer reported that Fireworks said an attacker gained entry to its development environment through a third-party cloud provider on October 23. The accessed data at Fireworks included resource metadata, billing information, model weights, and datasets for some customers, BleepingComputer said. Fireworks also had a separate incident in early October, when attackers reached environment variables holding API keys and tokens for a limited number of accounts.
What data was exposed
When Fireworks handed Notion the confirmed dataset on November 8, Notion found it contained Notion AI chat transcripts tied to the Insights feature. Insights was a tool that let workspace administrators see how their teams used AI inside Notion.
For most of the affected workspaces, Notion said the transcripts were unattributable. That means the records carried no labels identifying the author, the workspace they came from, or the results of any tool calls the AI made. Outside the content of the chat itself, there was nothing linking a transcript to a specific person or organization.
Notion also noted that the transcripts did not contain tool call results. Tool calls are commands an AI issues to fetch or write data, such as looking up a document or running a web search. Their absence limits how much context an attacker could piece together from the transcripts alone.
No other Notion features were affected, according to the company's advisory. Notion had also used Fireworks as a general inference provider to serve open-weight models inside the product, in addition to the Insights classification work. Both uses have now stopped.
What Notion did after confirming the scope
"We have contacted our customers who were affected by the Fireworks incident," Notion said in its security advisory.
Beyond customer outreach, Notion took two additional steps on November 12. It reported any exposed third-party API keys found inside affected user messages to the vendors that issued those keys. It also revoked Notion API keys that turned up in affected messages during its investigation, so those credentials can no longer be used.
The Insights feature has been turned off while the investigation continues. Notion said it will update its advisory as new information becomes available, and that this was not a breach of its own systems.
What affected customers should do now
Notion's guidance is clear: treat any data in the exposed transcripts as if it were fully public. If a workspace owner or team member typed passwords, API keys, or other credentials into Notion AI during the period covered by the incident, those credentials should be rotated right away.
Affected customers can review the specific data involved in the exposure. Notion said instructions were sent by email, and the data can be retrieved securely within the workspace itself.
For customers who have not received a notification, Notion said no immediate action is required. The company has not identified those workspaces as affected, though it added that its investigation is ongoing and it will reach out directly if that changes.
What is still not known
Several important questions remain open. Notion has not disclosed exactly how many workspaces were affected beyond its statement that the incident is impacting less than 1% of workspaces. The company has not described what the attackers did with any data they reached.
It is also not yet known whether any of the exposed chat transcripts have been misused. Notion said its investigation is ongoing, so there is no public confirmation that those behind the Fireworks breach have acted on what they found.
Why this matters for Notion users
Even users outside the affected group have reason to pay attention. AI productivity tools are woven into daily work, and those conversations often carry sensitive context: internal project names, draft strategies, and sometimes passwords or API keys copied in for a quick question.
This incident shows the downstream risk that comes with vendor relationships. Notion's own systems were not breached, but data shared with a third party for processing was still reached by attackers. The exposure path ran through Fireworks, not Notion, yet the data at risk belonged to Notion customers.
The practical takeaway is a simple policy: avoid pasting credentials into any AI chat interface, even one inside a trusted product. AI chat logs may travel to several systems, and a breach anywhere along that path can surface what was typed. Rotating any credentials that may have been shared this way removes a real risk at low cost.
Sources
- NotionFireworks AI incident
More from the network
Junious Media Network ↗Slow Horses Season 8 Will Be the Show's Last, Olivia Cooke Says
Olivia Cooke says she is filming what she believes is the final season of Slow Horses, though Apple has made no official confirmation.
junidaily.media ↗Glenn Close, Ridley Scott Receive Honorary Oscars at 17th Governors Awards
Five film veterans, including Glenn Close and Ridley Scott, were honored at the Academy's 17th Governors Awards, kicking off the race for the 2027 Oscars.
junidaily.media ↗California Piers Close as Waves Top 50 Feet This Weekend
Surf exceeding 50 feet battered San Mateo County Saturday, closing piers from Ventura to San Diego and triggering an evacuation warning near Capitola.
junidaily.media ↗Amanda Seyfried to Make SNL Hosting Debut on December 5
Amanda Seyfried hosts Saturday Night Live for the first time on Dec. 5, joined by Don Toliver, while her Elizabeth Holmes documentary drives broad buzz.
junidaily.media ↗