Notion AI Chats Exposed in Breach at AI Provider Fireworks
Notion has notified customers that their Notion AI chats were exposed in a breach at one of its vendors, Fireworks AI. Under 1% of workspaces are affected, but the chat content itself leaked; on 11 November Notion revoked API keys it found in affected messages.
Key facts
- Under 1% — of Notion workspaces affected
- 23 October — a third-party cloud provider hosting Fireworks R&D machines had a security incident; the attacker then reached Fireworks development machines
- 2 November — Fireworks alerted Notion; at first the dataset appeared to contain only synthetic test data
- 3 November — Notion disconnected its systems from Fireworks and deleted all of its data there
- 8 November — Fireworks confirmed the final dataset, which included real customer chats
- 9 November — Notion began contacting affected customers
Why this matters
- The supply chain is the real risk: your AI chats do not only live in the tool you use — they can travel to its sub-processor, and from there to criminals. Notion used Fireworks to serve open-weight models and to run classification jobs for Insights, a feature showing AI usage inside a workspace.
- “Unattributable” is not safe: most affected transcripts were not labelled with who wrote them or which workspace they came from, and did not include tool call results — but they still contain what users typed, including any passwords, API keys or internal information pasted into the chat.
- The attack is linked to LLMjacking: a financially motivated group that sells unauthorised access to large language models. Google Threat Intelligence Group and Mandiant warned about its rise on 2 November. Stolen model access can be resold, becoming another way to monetise a breach.
- There had already been an earlier intrusion in early October, which exposed stored secrets such as API keys. Fireworks has brought in Palo Alto Networks Unit 42 to help with the investigation.
- What to do: owners of affected workspaces should treat the data as exposed and rotate any passwords or API keys pasted into Notion AI during the incident window. If neither you nor your workspace owner has heard from Notion, no action is needed for now. Insights stays disabled, and Notion no longer serves open-weight models through Fireworks.
What it means for Hong Kong SMEs
For Hong Kong businesses using AI tools, the questions that matter are: where do my chats go? Do they reach a third-party model provider? Can I choose not to send them there? Staff pasting passwords, API keys and customer information into AI chats is one of the most common sources of exposure. FinityAI helps clients design agent permissions and data flows so these questions have answers, not guesses.
Sources
- Notebookcheck: Notion AI chats exposed in breach at AI provider Fireworks (12 Nov 2026)
- Notion security advisory (cited in the original)
This article is a FinityAI summary and commentary on a public report, and is not affiliated with Notion or Fireworks.
Related services
SRAA services · AI Agent security readiness · AI Agent workflow planning