Severity
Exposure
Date of incident
Samsung ChatGPT Secret Paste
confirmed
Tier -
A
Secrets crossed into model provider — no tokenization at agent ingress boundary.
What Happened
Employees treated chat agent as scratchpad; proprietary logic entered model provider retention boundary — governance gap at human–agent ingress.
Sumarry
Samsung restricted ChatGPT after staff pasted sensitive source code and meeting notes into consumer chat agent.
Detect
DLP + mask at ingress
What Happened
Trust Vault before any model API call
Blekline //
Mask PII/secrets at ingress before model API
Without agent boundary //
Ban consumer ChatGPT policy — shadow AI bypasses via personal accounts
