Notion Confirms AI Chat Transcripts Exposed in Fireworks AI Breach, Affecting Customer Workspaces

Notion has confirmed that a security incident at Fireworks AI exposed Notion AI chat transcripts linked to fewer than 1% of its customer workspaces.

Insights, a feature that helps teams understand their AI usage, used the affected conversations. Notion said its own systems were not breached and no other Notion features were affected.

The company has contacted affected customers and provided instructions for reviewing the impacted data securely inside their workspaces.

Customers who have not received a notification have not been identified as affected and need no action right now. However, Notion’s investigation remains open, and it will contact additional customers if it finds further exposure.

Notion AI Chat Transcripts Exposed

Fireworks first alerted Notion on the evening of November 2, 2026, after attackers accessed a small dataset. Notion’s initial review found only synthetic testing data, not customer information.

By the morning of November 3, Notion had disconnected its systems from Fireworks, deleted all data held there, and stopped communication with the provider.

The findings changed on November 6, when Fireworks warned that attackers might have accessed customer data. Notion launched another investigation, and Fireworks confirmed the new dataset on November 8.

That review identified customer chat transcripts processed for Insights, changing the company’s earlier assessment that no customer data was involved.

The disclosure follows the wider Fireworks AI data breach, which involved customer records, datasets, and incomplete model weights. Those broader findings should not be confused with Notion’s confirmed exposure, which concerns Insights chat transcripts.

Notion has not said that every data category affected at Fireworks was also exposed for its customers.

According to Notion’s security advisory, transcripts from the vast majority of affected workspaces were “unattributable.” They lacked labels identifying the author or workspace and did not contain tool call results.

Beyond the conversation text itself, Notion said there was no information linking these chats to users, workspaces, or companies.

That distinction does not mean the conversations were free of sensitive details. Information typed into a chat could still identify a person or business.

Separate coverage of exposed AI chat messages shows why conversation content matters even when the underlying AI model has not been compromised. These incidents have different causes, but both show the privacy risks surrounding stored chats.

Notion used Fireworks for Insights classification jobs and to run open-weight models inside its product. It has disabled Insights during the investigation and stopped serving those models through Fireworks.

Its AI settings documentation confirms that Insights is temporarily paused, with use cases and conversation counts unavailable.

Affected customers should treat the identified transcripts as exposed. Notion recommends changing passwords, API keys, and other credentials pasted into Notion AI during the incident timeframe.

Reviewing the supplied data is important because missing account labels cannot protect secrets written directly into a conversation. Further incident updates are available through Fireworks’ security bulletins.

Give your SOC the alert context to stops incidents quickly. Integrate Threat Intel from 16K SOCs.

The post Notion Confirms AI Chat Transcripts Exposed in Fireworks AI Breach, Affecting Customer Workspaces appeared first on Cyber Security News.