Hugging Face's logo

Spaces:
jonmabe
/
privacy-classifier-demo
Sleeping

App Files Community
privacy-classifier-demo / README.md
jonmabe's picture
jonmabe
Upload folder using huggingface_hub
0bcfec4 verified
preview code
|
raw
history blame contribute delete
1.11 kB

A newer version of the Gradio SDK is available: 6.9.0

Upgrade
metadata 
title: Privacy Classifier
emoji: 🔒
colorFrom: red
colorTo: green
sdk: gradio
sdk_version: 4.44.0
python_version: '3.11'
app_file: app.py
pinned: false
license: apache-2.0

Privacy Classifier Demo

Classify prompts to determine if they contain sensitive information that should stay local or if they're safe to send to cloud LLM services.

Classifications

  • 🔴 KEEP_LOCAL: Contains PII, sensitive data, or private information
  • 🟢 ALLOW_CLOUD: Safe to process with cloud-based AI services

Model

This demo uses jonmabe/privacy-classifier-electra, an ELECTRA-based classifier fine-tuned to detect sensitive information in prompts.

Use Cases

  • Privacy-aware prompt routing
  • Data loss prevention for LLM applications
  • Compliance with data protection regulations

Examples

Prompt Classification
"What is the capital of France?" ALLOW_CLOUD
"My SSN is 123-45-6789" KEEP_LOCAL
"Write me a poem about the ocean" ALLOW_CLOUD
"My password is hunter2" KEEP_LOCAL