Notion says a security incident at AI inference provider Fireworks AI may have exposed some Notion AI chats

Notion says a security incident at one of its vendors, Fireworks AI, may affect fewer than 1% of Notion workspaces. In its notice, Notion says some Notion AI chat transcripts used for its Insights feature may have been exposed. It describes them as "unattributable": not labeled with the author or workspace, and without tool call results such as page content. It says no other workspace data is affected. The Hacker News thread about the notice had 62 points and 18 comments by November 10.

Notion's account of the timeline:

• It learned of potential incidents at Fireworks earlier in November and shut down its connections to Fireworks. By November 3 it had disconnected its systems, deleted its data there and stopped sending traffic.
• On November 6, Fireworks told Notion that malicious actors may have accessed customer data. On November 8, Fireworks confirmed the new dataset for Notion.
• Notion used Fireworks for classification jobs behind Insights and as an inference provider for open-weight models. Insights is disabled during the investigation and Notion no longer serves open-weight models through Fireworks. It also halted training runs; it says participants in its early-access program had volunteered data to test task-specific models trained on Fireworks.

Notion has contacted the owners of affected workspaces, who can review the data under "Security incident review" in the desktop app or web. Enterprise owners get a CSV and an annotated CSV; other owners get a link to share with members. Notion recommends treating the data as exposed and rotating any passwords, API keys or other credentials that anyone pasted into Notion AI.