Notion, the workspace app many small businesses use for notes, wikis and project tracking, has confirmed that a security incident at one of its AI suppliers exposed Notion AI chat transcripts from a small share of workspaces. Notion's own systems were not breached, but the episode shows how anything typed into an AI assistant can end up with a company you have never heard of.
What happened
BleepingComputer reports that Notion used Fireworks AI, an AI infrastructure provider, to run some open-weight models inside Notion AI and to analyse AI conversations for a feature called Insights. Insights gives workspace administrators a view of how their teams use AI.
Fireworks AI has separately disclosed that an attacker got into its development environment through a third-party cloud provider in late October. According to Notion's security advisory, Fireworks first told Notion in early November that only synthetic test data had been accessed. Notion cut its connection to Fireworks, deleted its data from Fireworks' systems and stopped sending it traffic. Fireworks later told Notion that customer data may have been involved, and on 8 November it confirmed the final affected dataset. Notion began contacting affected customers the next day.
That dataset contained Notion AI chat transcripts collected for the Insights feature. Notion says that for the vast majority of affected workspaces the transcripts were "unattributable", meaning they were not labelled with the person or workspace they came from, and they did not include the results of tool calls. Insights has been switched off while the investigation continues, and Notion says it no longer serves open-weight models through Fireworks.
Who is affected
Notion says fewer than 1% of its workspaces are affected and that it has contacted those customers directly. If neither you nor your workspace owner has received a notice, Notion says it has not identified your workspace as affected and no action is needed for now. Its investigation is ongoing, and it will contact you if that changes.
Workspace owners who were notified can review the affected data in a desktop browser through the help menu in the sidebar, under Security incident review. From there they can download a list of affected members and share a link so each member can see their own affected data. Owners with a verified domain can also download the affected content itself.
Why it matters for Australian businesses
Even if your business is not in the affected group, this incident is worth taking seriously. Staff often paste client emails, contract wording, financial figures and sometimes passwords or API keys into AI assistants to get a quick answer. That text can be stored, processed or analysed by subcontractors behind the product you actually signed up for.
BleepingComputer notes that Fireworks had committed to zero data retention for Notion data, yet transcripts were still exposed. Contract promises from your supplier's suppliers are not the same as control. If personal information about your customers is in those chats, you still need to think about your obligations under Australian privacy law and whether the Office of the Australian Information Commissioner's notifiable data breaches scheme applies to you.
This follows other recent supplier incidents we have covered, and it fits the same pattern: the weak point is often a provider one or two steps removed from your business.
What to do now
- Check for a Notion notice. Ask whoever owns your Notion workspace to search their inbox for a message from Notion about the Fireworks AI incident. If one arrived, open Security incident review on a desktop browser and download the affected member list.
- Review what was exposed. Share the member link with affected staff and ask them to check their own transcripts for client details, passwords, keys or other sensitive information. Notion recommends treating this data as exposed.
- Rotate anything sensitive. If passwords, API keys, tokens or other credentials were pasted into Notion AI, change them now and turn on multifactor authentication for those accounts where available.
- Assess privacy obligations. If the chats contained personal information about customers or staff, work through whether the incident is likely to cause serious harm and whether you need to notify, using the OAIC guidance linked above.
- Set a simple AI use rule. Agree in writing which AI tools staff may use for work and what must never be pasted into them, such as passwords, keys, bank details and identity documents. Store secrets in a password manager instead.
- List your AI suppliers. For each AI tool in use, note who the vendor is, whether it uses sub-processors and how it notifies you about incidents. Keep the Essential Eight basics in place so a leaked credential is less useful to an attacker.
How MSP Logic can help
AI tools are now part of everyday work, and most businesses have no record of which ones staff use or what goes into them. MSP Logic helps Australian businesses map AI and cloud tools that handle company data, write a practical AI use policy, and rotate credentials safely when a supplier reports a breach. See our cybersecurity services for small business, try the Free IT Health Check, or get in touch for a quick review.
Source: BleepingComputer – Notion says Fireworks breach exposed customer AI chat transcripts. This article is MSP Logic's own summary and advice; read the original report for full details.