README.md

metadata

title: Privacy Filter Flask Demo
emoji: 🔒
colorFrom: blue
colorTo: purple
sdk: docker
pinned: false
license: apache-2.0
short_description: Flask demo for OpenAI Privacy Filter PII detection model

🔒 OpenAI Privacy Filter Flask Demo

A Flask-based web interface to demonstrate OpenAI's Privacy Filter model for detecting and masking Personally Identifiable Information (PII) in text.

Model

openai/privacy-filter - A bidirectional token-classification model trained specifically for PII detection.

Detects 8 Privacy Categories

Category	Description	Example
private_person	Names of individuals	"Alice Smith", "John Doe"
private_email	Email addresses	"alice@example.com"
private_phone	Phone numbers	"(555) 123-4567"
private_address	Physical addresses	"123 Main Street, NYC"
account_number	Account/ID numbers	"Acct: 12345678"
secret	Secrets/credentials	Passwords, API keys, tokens
private_url	Personal URLs	Private links
private_date	Personal dates	Birthdays, dates of birth

Usage

Enter any text containing PII in the web interface and the model will:

1. Highlight detected PII entities with color coding
2. Label each entity by type
3. Show confidence scores for each detection

Technical Details

• Framework: Flask
• Model: openai/privacy-filter (1.5B params, 50M active)
• License: Apache 2.0
• Context Window: 128,000 tokens
• Architecture: Bidirectional token classifier with Banded Attention

Links

• Model Card
• OpenAI Privacy Filter GitHub
• Model Paper/Card PDF