Notion AI chat transcripts exposed in Fireworks AI breach affecting under 1% of workspaces

Notion has informed a small number of customers that their information was compromised in a data breach involving Fireworks AI, a cloud platform used to run and fine-tune open‑source AI models. SecurityWeek's coverage confirms that Fireworks AI was hacked twice in October: first through compromised internal credentials, and then via a third‑party cloud provider hosting Fireworks R&D machines.

The second incident occurred on 24 October, when attackers executed an unauthorized workload that led to the exposure of customer data. Fireworks notified Notion on 2 November that Notion AI data had been accessed. Prior to the incident, Notion used Fireworks as a general inference provider for delivering open‑weight models, according to SecurityWeek.

Investigators determined that hackers accessed Notion AI chat transcripts used for the Insights feature, which assists customers in understanding AI usage. Fewer than 1% of Notion workspaces were affected, and most transcripts were unattributable to a specific customer or workspace. Notion says there is no information tying the chat transcripts to particular workspaces, users, or companies beyond the transcripts themselves.

Affected customers have been contacted with instructions on securely retrieving and reviewing the data in their workspaces; Notion did not disclose how many customers were affected. Notion advises treating any exposed data as such and to rotate credentials if sensitive information, passwords, API keys, or other credentials were pasted into Notion AI during the incident. SecurityWeek has sought further information from Notion and will update if responses are received.