A transcript can contain names, private details, or credentials even when the file has no name.
On November 9, Notion disclosed that an incident at Fireworks AI affected AI Insights transcripts from fewer than 1% of Notion workspaces. Notion says most transcripts in the dataset did not carry user or workspace identifiers. The text could still contain details that identify people or organizations. Notion says the dataset did not include tool call results.
What the companies reported
Notion says Fireworks initially reported on November 2 that the dataset appeared to contain synthetic data. Notion disconnected the integration on November 3. Fireworks flagged possible customer data on November 6, and Notion says it confirmed the dataset on November 8. Fireworks published a separate bulletin describing its investigation and the types of data involved at its end.
Notion says it disabled Insights and is notifying affected users. The vendor’s bulletin describes its broader incident, so use Notion’s advisory for the scope of the exposure in Notion.
What to check if Notion notified you
Open Notion directly and follow its instructions to review the affected content in the workspace. Look for any credentials or sensitive details people put in the transcripts. If you find a token or password in affected text, revoke or rotate it at the service that issued it. For a work account, tell the person responsible for that service which credential was exposed.
Review what is actually in the affected transcript before deciding which other systems or people need to be notified.
Abdessalam Alaoui
I'm Abdessalam Alaoui, founder of Musthave.AI. I write practical guides to AI models, tools, generative media.
