🛡 SafeType

AI-Powered Parental Monitoring for Android

Protect conversations before they become crises.

Monitor WhatsApp, Instagram, Snapchat and more with our custom-trained deep learning model. 10-category detection flags bullying, hate speech, threats, and other harmful content in real time.

Setup Guide Parent Dashboard

10+ apps monitored · Real-time alerts · Custom DistilBERT model · Transparent monitoring

How SafeType works

1

Install SafeType on your child's Android device. It captures messages via keyboard, notifications, and screen reading.

→

2

Messages are sent to our cloud API where a custom DistilBERT model classifies content across 10 safety categories.

→

3

View messages, detailed analysis, and threat alerts on your parent dashboard. Click any message for full breakdown.

Why parents choose SafeType

10-Category Detection

Our custom model detects racism, sexism, cyberbullying, hate speech, threats, toxicity, profanity, implicit hate, sarcasm, and clean content with probability scores for each.

Cross-App Visibility

One dashboard for WhatsApp, Instagram, Snapchat, Messages and more. No need to check each app separately.

Educational Safety Guide

Each detection comes with actionable advice for parents and age-appropriate explanations for children about what the content means and how to respond.

Real-Time Sync

Messages appear on your dashboard within seconds. Click any message for detailed 10-category probability breakdown and safety guidance.

Safety & Transparency

SafeType is designed for ethical, above-board monitoring.

Child-visible notification: A persistent notice on the device says “SafeType Parental Monitor is active.” Children know they are being monitored.
Not covert: We believe parents and kids benefit from clear expectations. No hidden tracking.
For safety, not surveillance: Built to protect kids from harm, not to spy. Use responsibly.

Built at Duke University

SafeType is a Deep Learning research project at Duke University. We developed and deployed a custom DistilBERT model trained on real-world harmful content data, replacing third-party APIs with our own model for transparent, controllable safety classification.

Our Model

A fine-tuned DistilBERT classifier that detects 10 categories of harmful content: racism, sexism, profanity, cyberbullying, toxicity, hate speech, implicit hate, threats, sarcasm, and clean messages.

Calibration

Post-processing calibration reduces false positives while maintaining sensitivity to genuinely harmful content. Severity-tiered thresholds treat categories differently based on their real-world impact.

Integration

The model runs on our deployed API, receiving messages from the Android app and returning real-time classification results with full probability distributions.

Get started

Install the app on your child's device, then sign in to your dashboard.

Setup Guide Parent Login

Common questions

What AI model do you use?: We use a custom fine-tuned DistilBERT model trained at Duke University. It classifies text into 10 categories with calibrated confidence scores, running on our own deployed API.
Is my child's data private?: Messages are stored in a secure database and used only for safety analysis. The system is designed for a single family; we do not sell or share data.
What about false positives?: Our model includes post-processing calibration to reduce false positives. Parents always see the full message, probability breakdown, and flag reason to judge for themselves.
Does it work on iPhone?: No. SafeType requires Android (keyboard and accessibility features). iOS does not allow equivalent monitoring without enterprise MDM.
How do I get the app?: Follow the Setup Guide to download the APK from our secure link and install it on your child's Android device.

Setup Guide

Get SafeType running on your child's device in under 10 minutes. You'll need physical access to their Android phone.

1

Download the APK

Download the SafeType installer from our secure link. The file is approximately 15 MB.

Download from Google Drive

2

Enable installation from unknown sources

On the child's device, go to Settings → Security → Install unknown apps and allow your file manager or browser to install APKs.

3

Install and grant permissions

Open the downloaded APK and install SafeType. The app will request three permissions:

Accessibility Service — reads on-screen messages from chat apps
Notification Access — captures incoming message notifications
Input Method — enables the SafeType keyboard to capture outgoing text

All three must be enabled for full monitoring coverage.

4

Set SafeType as the default keyboard

Go to Settings → System → Languages & input → Default keyboard and select SafeType.

5

Verify on your dashboard

Send a test message from the child's device. Within 10–15 seconds, it should appear on your parent dashboard. The device status will show as "Online".

Transparency: SafeType shows a persistent notification on the child's device that says "SafeType Parental Monitor is active." This is intentional — children should know they are being monitored.

Parent Dashboard

0 Messages today

0 Threats detected

0 Apps monitored

-- Last sync

No messages yet. Waiting for device sync...

Detection Result

Click a message to see detailed analysis