Notion AI chats exposed through a vendor breach: the lesson for small teams
A breach at AI supplier Fireworks AI exposed some Notion AI chat transcripts this month. What happened, and what your staff should never paste into AI tools.
Notion, the note-taking and workspace app used by many small teams, has disclosed that some Notion AI chat conversations were exposed after a breach at one of its AI suppliers. Notion's own systems were not breached. The incident is a clear example of a risk most small businesses have not thought through: when staff type into an AI assistant, that text can travel to companies you have never heard of.
What Notion has said
According to Notion's security advisory, its vendor Fireworks AI, which it used to run some AI models and to classify conversations for an analytics feature called Insights, told Notion on November 2, 2026 that a small set of data had been accessed. At first this was described as synthetic test data. Notion disconnected from Fireworks by November 3 and deleted its data there on November 4. Fireworks later said customer data may have been accessed, and confirmed the affected data set on November 8. Notion began contacting affected customers on November 9.
The exposed data was Notion AI chat transcripts from fewer than 1% of workspaces. Notion says most of them could not be tied to a named person or workspace, and that no other Notion features were affected. SC Media also reported the change from "test data only" to confirmed customer data. Notion has disabled Insights, stopped using Fireworks, and on November 11 revoked Notion API keys it found pasted into affected chats. Its advice to affected customers is to treat the data as exposed and change any password, API key or credential that was pasted into Notion AI.
The lesson for small businesses
This was not careless security at a tiny startup. It shows how AI features work: the app you use often passes your text to other providers that run the models or analyze the results. Each one is another place your data sits, and another company that can be breached. The same is true of many AI assistants built into email, document, CRM and accounting tools.
The part a business can control is what goes into those chats in the first place. The detail that Notion had to revoke API keys found in conversations tells you people paste secrets into AI tools, usually without a second thought, while asking for help with a script or an error message.
What to do this week
- If you use Notion, check your email and the admin notices in your workspace. If you were contacted, change any credentials that were pasted into Notion AI, starting with anything that gives access to money, customer data or admin accounts.
- List the AI tools your team actually uses, including AI features switched on inside software you already pay for. Note who uses them and for what.
- Write a one-page AI rule. At minimum: never paste passwords, API keys, recovery codes, card or bank details, Social Security numbers or client files into an AI chat, approved tool or not.
- Prefer business accounts over personal ones. Company-managed AI accounts let you see who has access, remove leavers and review data settings. Personal free accounts give you none of that.
- Keep secrets in one place. A business password manager makes it far less tempting to paste a key into a chat window.
How we can help
We can help you review which outside services hold your data and set sensible rules for staff, as part of our cybersecurity and data loss prevention work. Call (213) 986-4855.