On-Device AI: Why Your Content Never Leaves Your Phone

The centerpiece of Remain Faithful's privacy model is on-device AI classification. Here's what that means in practice, and why it matters.

The ReplayKit Extension

iOS's ReplayKit framework creates a separate extension process for screen recording. The extension runs all classification locally — on the device's Neural Engine — and uploads only alert metadata when a frame is flagged. That metadata contains the alert category, severity level, a brief static description, and a timestamp. No screen content, OCR text, or screenshots ever leave your device.

That's not a policy commitment. Anyone can verify it by reading the open-source code.

Two Tiers of On-Device Classification

Tier 1: URL Blocklist + Keyword Matching

When a browser is detected, visible URLs are checked against a local blocklist of known adult domains. Visible text is pattern-matched against regex rules for explicit content categories. This is fast, deterministic, and 100% local.

Tier 2: Apple Vision + SensitiveContentAnalysis + Local Classifier

Apple provides two relevant frameworks: Vision OCR (which extracts text from screen frames) and SensitiveContentAnalysis (which detects nudity and explicit images). A local keyword classifier then scores the OCR output across content categories. All three run on the device's Neural Engine, the dedicated AI chip in modern iPhones. No server involved at any stage.

Why This Architecture Matters

A lot of accountability tools say they're "private." What that usually means is that they've made a policy commitment not to look at your data.

Policy commitments can change. Code is harder to change unnoticed. By running classification entirely on-device and making the codebase open source, we've created a system where the privacy guarantee doesn't depend on trusting us — it depends on reading the code.

We think that's a more honest form of privacy.