Notion AI Chat Transcripts Accessed Through an AI Vendor's Breach (Fewer Than 1% of Workspaces): What Notified Workspaces Should Do

According to Notion's official security advisory, a vendor it used for AI processing suffered a security incident, and malicious actors accessed data that included Notion AI chat transcripts used for the Insights feature. The company says fewer than 1% of workspaces are affected and that this was not a breach of Notion's own systems.

The company says it has contacted the affected workspaces. If you have not heard from it, it has not identified your workspace as affected and no action is needed right now. The investigation is ongoing, and it will contact you directly if your workspace is found to be affected. Check your spam folder too.

In the company's account, Notion AI chat transcripts used for the Insights feature. For the vast majority of impacted workspaces the chats were not labeled with who wrote them or which workspace they came from, and they contained no tool call results. Apart from the chat text itself, the company says nothing ties a chat to a workspace, user or company. A chat transcript includes any input to or output from the model that was written to the chat. AI Meeting Notes audio, transcripts and summaries were not affected unless used as input in a chat with Notion AI. The first dataset reported to it was Notion AI synthetic testing data and contained no customer data.

The company recommends that people in affected workspaces who pasted sensitive information, passwords, API keys or other credentials into Notion AI during the incident timeframe rotate them. The advisory does not give specific dates for that timeframe. If you were notified, follow the notice; if unsure, rotating any credential you have ever pasted into Notion AI is the safe course (this site's view). In its November 11 update the company says it revoked exposed Notion API keys found in affected user messages. It does not say it revoked any other credentials, so rotate those yourself (this site's view).

The company says this was not a breach of its systems. At its AI vendor, malicious actors accessed a small set of data and later data that may have been customer data. The advisory does not say how the vendor was compromised. The vendor's own official statements are summarized in a separate article on this site.

According to Notion's help page, the vendor had given a contractual commitment to zero data retention for Notion's use of its platform. During its investigation, the company learned that data was retained from the vendor endpoints it used. The company also explains that features that need stored requests and responses, such as chat history and Insights, are not covered by the enterprise zero data retention commitment. It says it is auditing critical vendors, such as inference providers and data processors, for data minimization and zero data retention practices.

According to Notion's help page, owners of affected workspaces open Security incident review from ? in the sidebar. It requires the latest version of Notion in a desktop browser and is not available on mobile. Owners can download a CSV of affected members (names, emails, membership status, user IDs and the count of affected rows) and copy a link so members can review their own data. Owners of workspaces with a verified domain can also download CSVs of the affected content, in which conversations that may contain sensitive information are flagged. Owners can remove members' access to their own affected content; this does not delete any data.

According to the advisory, the Insights feature is disabled while the incident is investigated, and open-weight models are no longer served through this vendor. The company also says that as of November 3 it stopped sending data and traffic to the vendor and halted all training runs there. The advisory does not say whether the rest of Notion's AI features are unchanged, and it gives no date for Insights to return.