Family Digital Safety Copilot iOS / Android

Family Digital Safety Copilot is a mobile app (iOS and Android) that helps parents and guardians protect children and teens online without resorting to invasive, trust-breaking surveillance. Instead of simply logging and reporting everything a child does, the app acts as an AI “copilot” that watches for real risk signals — grooming attempts, cyberbullying, self-harm ideation, exposure to explicit content, scams — and intervenes early, while giving the child age-appropriate autonomy and the parent clear, actionable insight.
The core idea: less “spy on your kid,” more “co-pilot that flags danger and coaches both sides.”

3. How It Works — End to End

Step 1: Family Account Setup

• A parent creates a Family Hub account and adds each child as a profile (with age, grade, and risk-sensitivity settings).

• Each child installs a lightweight companion app or enables OS-level permissions (Screen Time API on iOS, Digital Wellbeing/Accessibility API on Android) that let the Copilot observe signals, not raw content by default.

• Setup includes an honest, age-appropriate conversation flow: the child is told what is and isn’t monitored, building transparency instead of secrecy.

Step 2: On-Device Signal Processing

• Most analysis happens on-device for privacy: message metadata, app usage patterns, screen time, and (with permission) message/content scanning run through a local lightweight classifier.

• The classifier looks for patterns associated with:

• Grooming or predatory contact (sudden secrecy requests, off-platform contact requests, escalating flattery from unknown adults)

• Cyberbullying or harassment

• Signs of distress, self-harm ideation, or disordered eating in text

• Exposure to explicit or violent content

• Scam/phishing attempts targeting the child

• Only when a pattern crosses a risk threshold is anything sent to the cloud AI layer for deeper contextual review — raw everyday chats are not uploaded or stored.

Step 3: Cloud AI Escalation Layer

• Flagged signals are sent (encrypted, anonymized where possible) to a cloud model that provides deeper contextual judgment: is this actually concerning, or a false positive (e.g., song lyrics, a joke between friends)?

• The AI assigns a risk tier: Low (log only), Medium (nudge/coach the child), High (alert parent), Critical (alert parent immediately + provide crisis resources).

Step 4: Response — Tiered by Severity

• Low risk: Nothing shown to parent; may be used to fine-tune the child’s personal “digital wellbeing” trends.

• Medium risk: The child gets an in-the-moment nudge — e.g., “This message seems like it’s trying to get you to keep a secret from your parents. Want tips on how to respond?” This builds the child’s own judgment instead of just reporting on them.

• High risk: Parent receives a plain-language alert (“Your child received a message from an unknown adult asking to move the conversation to a private app”) with suggested next steps, not raw transcripts, preserving some of the child’s privacy while keeping parents informed.

• Critical risk (e.g., explicit predatory contact, self-harm indicators): Immediate parent notification with full relevant context, plus in-app links to appropriate hotlines/resources and guidance on next steps.

Step 5: Parent Dashboard

• A Family Hub dashboard shows:

• Screen time and app usage trends

• Risk alerts timeline (with severity tags)

• Digital wellbeing score (sleep-hours device use, social comparison exposure, etc.)

• Conversation starters — AI-generated, non-accusatory scripts to help parents talk to kids about what was flagged

• Parents can adjust sensitivity per child, per app, and per age milestone (a 16-year-old gets more autonomy than a 9-year-old by default).

Step 6: Ongoing Coaching Loop

• The system doesn’t just flag problems — it teaches. Kids get bite-sized, non-preachy media-literacy and safety tips triggered by real situations they encounter (contextual, not generic lessons).

• Parents get periodic guidance on age-appropriate digital independence, so the app’s restrictiveness decreases over time as trust and maturity increase — designed to be phased out, not permanent surveillance.

5. Privacy & Trust Principles

1. Transparency over secrecy — the child always knows the app is active.

2. Signal-based, not transcript-based — parents see summaries and risk flags, not a full surveillance feed, except in critical safety situations.

3. On-device-first processing — minimizes data leaving the child’s phone.

4. Data minimization & encryption — anything sent to the cloud is encrypted and retained only as long as needed for safety review.

5. Designed to fade out — the product’s stated goal is to build the child’s own judgment, reducing reliance on monitoring as they mature.