Skip to content
AI news · Privacy & rules

Some Notion AI chats leaked through a supplier's breach, a reminder that your AI tools have suppliers too

By , Founder of Adaptd ·

Notion says fewer than 1% of workspaces may have had Notion AI chat transcripts exposed in a security breach at Fireworks AI, a company it used to run some AI features.

The short version

  • Some Notion AI chat transcripts were exposed in a breach at Fireworks AI, a company Notion used to run some AI features.
  • Notion says fewer than 1% of workspaces may be affected, and it has contacted the owners of those workspaces.
  • If anyone pasted passwords or API keys into Notion AI, change them now, whether or not you've heard from Notion.

You chose Notion. You probably didn't choose Fireworks AI. That's how a lot of AI security news works now.

What happened to Notion AI chats?

Some Notion AI chat transcripts were exposed in a security incident at Fireworks AI, a company Notion used for its Insights feature and to run open-weight AI models in its product. Fireworks told Notion on 6 November 2026 that attackers may have accessed customer data, and confirmed the dataset on 8 November.

Notion says the chats weren't labelled with who wrote them or which workspace they came from, and didn't include page content Notion AI pulled in. Notion cut its connection to Fireworks by 3 November, deleted its data there and switched Insights off.

Who is affected?

Notion says fewer than 1% of workspaces may be affected, and it has contacted the owners of those workspaces. If neither you nor your workspace owner has heard from Notion, Notion says no action is needed.

Affected owners can review the exposed chats on a desktop browser under the sidebar's ? menu, then Security incident review. Enterprise owners get a CSV; owners on other plans get a link to share with affected members.

What should your business do about it?

Your business should change any password, API key or login that someone pasted into Notion AI, because Notion advises treating the exposed chats as public. Ask the team directly; nobody remembers what they pasted at 4pm on a Tuesday.

My advice: find out which companies your AI tools pass your data to. Most publish a list of these suppliers, usually called subprocessors, on their trust or privacy page. You don't need to read every line. It's like a builder's subcontractors: you just want to know who's coming onto the site.

And keep passwords out of chat boxes altogether. A password manager is a much better place for them than an AI chat history.

Wondering what this means for your business?

That's exactly what the discovery call is for.