Skip to content
Safety & Trust

We don't trust "engagement AI" with our kids.
So neither should you.

We started with safety before we wrote a single feature. Here's what that actually looks like. Here's exactly how.

The 4 Layers

How a harmful message gets stopped.

Every message your child types goes through four independent safety checkpoints before NeoBuddi responds. No single layer is the whole defence. If one misses, the next catches it.

Layer 1

Every message is filtered first

A local screen catches violence, profanity, and crisis language before the AI ever sees it.

Layer 2

The AI itself refuses harmful content

Tuned to redirect toward curiosity, not shame. When it catches something concerning, it pulls back.

Layer 3

Every generated image is checked

Educational only. Nothing scary, nothing suggestive. A safety prompt evaluates each image before your kid sees it.

Layer 4

You in the driver's seat

PIN-protected dashboard. Full transcripts. Safety alerts in real time. One-tap export of your child's data.

The 4th layer is you.

You can read every message. Set time limits. Get alerts the moment anything concerning happens.

COPPA Compliant
GDPR-ready
Data never sold
No open web
Our Principles

Promises we put in code, not just in a policy PDF.

1. No engagement hacks. Ever.

No streaks. No infinite scroll. No push notifications designed to pull your child back. When they're done learning, NeoBuddi says goodbye. Not "just one more."

2. A walled garden, not the open web.

NeoBuddi cannot be prompted into an open web search. The only "web" access is a curated, age-appropriate fact-check layer in Premium, with the same moderation stack.

3. Your child is never the product.

We never sell data. We never use child conversations to train third-party models. All child identifiers are anonymized before any analytics event leaves the device.

4. Real transparency for parents.

Full conversation transcripts. Every generated image inspectable. Safety alerts in real time. PIN-protected access. One-tap export of all your child's data on demand.

5. Refusing to be cute about hard stuff.

If NeoBuddi detects distress, self-harm ideation, or abuse cues, it ends the "tutor" conversation and surfaces a parent alert plus crisis resources immediately.

Read the full privacy policy