Back to all lessons
Awareness Lessons
2 days ago

Third-Party AI Vendor Breach Exposes Notion Customer Chat Data

Written by AI from the linked story. Not reviewed by an editor. How this works

Notion's customer data was compromised not through a direct attack on Notion itself, but through a breach of Fireworks AI, a third-party vendor handling AI chat transcripts for the Insights feature. This is a classic supply chain security failure: sensitive customer data was processed and stored by an external platform that became the attack surface. The incident highlights the risk of extending trust implicitly to vendors who handle personal or sensitive data on your behalf. Even when a company's own security posture is strong, a weaker link in the vendor chain can expose customer data at scale. Organizations must treat third-party data processors as extensions of their own security perimeter and hold them to equivalent standards.

Tactical Insight

Immediate actions

  • Notify all potentially affected customers promptly and advise them to rotate any exposed credentials or tokens.
  • Conduct an emergency review of what data each active third-party AI or analytics vendor currently holds on your behalf.

Long-term improvements

  • Establish a formal Third-Party Risk Management (TPRM) program that requires vendors to meet defined security standards before onboarding.
  • Apply data minimization principles so that vendors only receive the minimum data necessary to deliver their service.
  • Include contractual breach notification SLAs and audit rights in all vendor agreements covering personal or sensitive data.

Detection measures

  • Implement continuous monitoring of data flows to third-party vendors to detect anomalous access or exfiltration.
  • Require vendors to provide timely security incident notifications and validate their incident response capabilities during onboarding.

Back this up with a control

All controls and vendors

Training helps people spot this. These controls make the attack fail even on the day someone does not.