AI document processing without leaking personal data

The scenario finds personal data in the document, replaces it with tokens, sends the anonymized text to AI, and restores the real values in the finished answer.

  • Only anonymized text reaches the model
  • Real values stay in your Postgres, not with the AI
  • If masking fails, processing is blocked
  • Every step is written to an audit log
Webhook
Set it upMade with Scriptera