aavhawkeye/privacy-gateway-training-data

Team Red Privacy Gateway Training Data

This dataset contains the current Team Red privacy-gateway training splits.

What It Trains

The privacy gateway is the local layer that:

• converts raw scan artifacts into reasoning-preserving sanitized packets
• replaces sensitive identifiers with stable placeholders
• emits typed hints for downstream reasoning
• refuses reverse-lookup or reveal requests

Why It Exists

Team Red keeps institution-specific identifiers local before any hosted reasoning step.

The companion adapter is published here:

• https://huggingface.co/aavhawkeye/privacy-gateway-adapter

Access Model

This dataset is public with gated manual review enabled.

That means the existence of the dataset is visible, but access requests must be manually approved.

Splits

• train
• validation
• test