Last updated: July 2026
FraudVerse AI processes the transaction/account data you upload or connect (e.g. via CSV upload) solely to compute fraud risk scores, SHAP explanations, and network/graph analysis. PII fields are tokenized at rest using a server-side secret before storage.
We do not sell your data. We do not use uploaded transaction data to train models outside this Platform. CSV previews on the Upload page are parsed client-side — no backend call is made until you explicitly submit for scoring.
The AI Assistant sends your chat message (and recent conversation context) to a third-party LLM API to generate a response. Avoid pasting sensitive account numbers, passwords, or personal identifiers into the chat.
In this demonstration deployment, uploaded and processed data is retained only for the duration needed to serve the current session/pipeline run, unless explicitly configured otherwise for a production deployment.
Access is logged via an audit-logging layer (who accessed which account, when). PII tokenization keys must be set via environment configuration before any real deployment — development builds use a fixed key for local testing only.
If this Platform is deployed for your institution, you may request details on what data is stored and request its deletion, subject to your organization's regulatory retention obligations (e.g. STR/AML record-keeping requirements).
This policy may be updated periodically to reflect changes in the Platform.