Local, AI-powered PII redaction for macOS

Built with OpenMed, MLX-Swift, and Apple Vision

⚠️ If macOS Gatekeeper blocks the app (it's not signed with a developer certificate), bypass it by running:

xattr -rd com.apple.quarantine /Applications/HideMyData.app

Install

Grab the latest .dmg from the Releases page, or build from source - see Build below.

demo.mp4

Features

• Fully local — model runs on-device, nothing ever leaves your machine
• PDF and image input — both formats share the same detection and redaction pipeline
• OCR — Apple Vision handles scanned PDFs, images, and rescues PDFs whose embedded fonts hide text from selection
• AI detection — OpenAI privacy-filter (MLX 8-bit) catches names, emails, phones, addresses, dates, IDs in context
• Manually maintained regrex — IBAN, SSN, Personal identifiers, MAC, IPv4/v6, JWT, API keys, crypto wallets and more to come
• Two redaction styles — solid black or frosted glass blur
• Manual editing — add or remove redaction rectangles by hand before saving
• Permanent on save — pages are rasterized and rebuilt - the original text and glyphs are gone, not just hidden

Requirements

• macOS 26 or later
• Apple Silicon (the MLX backend does not run on Intel)

Build

open HideMyData.xcodeproj
# build & run via Xcode (⌘R)

On first launch, you will be prompted to download the model (~1.5 GB) from Hugging Face into ~/Library/Application Support/HideMyData/ModelCache/.

Tech Stack

Swift 6, SwiftUI, MLX-Swift, Apple Vision, PDFKit, OpenMedKit

License

GPL-3.0