Simplify your safety strategy with one platform that respects trust.
Create a safer digital future with a privacy-first solution families and schools trust.
Open-source at heart—SafeCircle empowers everyone to build a safer digital world together.

SafeCircle AI Playground Is Now Open Source

Horizon
Horizon

A year of late nights, thousands of lines of code, and an unwavering belief that AI could make the digital world safer. Today, we're sharing the result a free, fully offline AI app for Android, built with a purpose.

One Year in the Making: SafeCircle AI Playground Is Now Open Source


One year ago, this was nothing more than an idea.

The premise was simple but felt enormous: what if AI could make the digital world safer for children? Not as an abstract promise, but as something real, running on the device already in a parent's pocket no subscription, no server, no data leaving the phone.

That idea became an obsession. Then it became code. Then it became this.


Introducing SafeCircle AI Playground


AI Playground is a free, open-source Android application that runs powerful AI models entirely on-device. No internet required. No account needed. No data ever leaves your phone.

It is built on Google AI Edge and powered by our own in-house Horizon model family fine-tuned, optimized, and designed around SafeCircle's mission.


What it does


  • Safety Detection: Analyze conversations in real time to surface risks like grooming, bullying, and exploitation, using our Horizon Mobile and Horizon Edge models
  • LLM Chat: Multi-turn conversations with an on-device language model, including the Horizon AI persona
  • Agent Chat: Agentic task execution with 12 built-in skills and support for MCP tools
  • Prompt Lab: Single-turn prompting and experimentation
  • Benchmark: Measure model latency and throughput directly on your hardware

Everything runs locally. GPU, CPU, or NPU your device, your choice.

Why we built this


At SafeCircle, we believe every child deserves a safer digital future.

Online dangers predators, exploitation, harmful content are not theoretical. They are happening right now, to real families. Existing tools either require cloud connectivity (meaning sensitive conversations leave the device) or are too complex for everyday use.

We wanted to build something different: AI that protects privacy while protecting people. The two are not in conflict. They are the same goal.

AI Playground is the foundation for that. It lets families, researchers, developers, and organizations explore what on-device safety AI can look like and build on top of it.


Fully offline. Completely free. Open source.


The app is available today on Android (API 30+) with no paywalls, no usage limits, and no telemetry.

The full source code is on GitHub under the Apache 2.0 license. We want the community to audit it, improve it, fork it, and push it further.

If our work helps even one child avoid becoming a victim, every late night and every setback will have been worth it.


Thank you


To everyone who followed our progress, tested early versions, shared feedback, spread the word, or simply believed in this vision before there was anything to show thank you.

Your patience and trust helped turn an idea into something real, and kept this project moving when the road was hard.

This is only the beginning.


Tomas, Founder & Developer, SafeCircle


More info