📘 The Crumbforest: Local‑First Child-Safe Learning Infrastructures

A UNICEF & Save the Children–Aligned Policy Brief
Version 1.0 ¡ February 2026

1. Executive Summary

The Crumbforest is a local-first, offline-capable learning infrastructure designed specifically for environments where children’s rights, safety, autonomy and privacy must be protected above all other considerations.

It was developed to serve:
- Schools without stable internet access
- Refugee and displacement camps
- Rural and low‑resource communities
- Child protection contexts where surveillance risks are high
- Crisis and conflict settings where digital services are unsafe

The system uses low-cost hardware (Raspberry Pi, ESP32), auditable open-source software, and local knowledge databases to enable learning without exposing children to:
- online tracking
- data extraction
- commercial profiling
- surveillance
- identity risks

The Crumbforest is not a platform. It is infrastructure that communities can host, inspect, reset, replicate and own.

It aligns with:
- UN Convention on the Rights of the Child (CRC)
- UNHCR Child Protection Minimum Standards
- Save the Children Digital Safeguarding Guidelines
- UNICEF Principles for AI for Children
- INEE Minimum Standards for Education

It is based on one core idea:

Children’s learning data should never leave the room they are learning in.

2. Rationale: Why Local‑First Matters

Digital education systems often rely on cloud platforms, centralized identity systems, continuous connectivity, and data harvesting by vendors. In many humanitarian and development contexts, this model is unsafe or impossible.

2.1 Safety

Children’s identity data stored externally can expose them to exploitation, misidentification, political risk, profiling, and coercion. Local-first systems remove these risks by eliminating external data flows.

2.2 Equity

Connectivity-based platforms exclude remote rural learners, displaced children, and emergency environments lacking stable power/internet. A local system ensures universal access, regardless of ISP availability.

2.3 Accountability

Local systems are:
- auditable
- transparent
- inspectable by NGOs and communities
- independent of commercial vendors

This supports sovereignty, trust, and long-term sustainability.

3. System Overview

3.1 Hardware

  • Raspberry Pi 4/5 (learning node)
  • ESP32 microcontrollers (engagement + feedback devices)
  • Local server (Debian-based, renewable-energy compatible)

All devices cost between €5 and €70, making them suitable for scale in low-resource contexts.

3.2 Software

  • Debian GNU/Linux (stable, secure, community-governed)
  • PostgreSQL + pgvector (local knowledge storage)
  • Gitea (local repository)
  • WireGuard (secure optional mesh)
  • Ollama (offline AI inference)
  • Go-based API (single binary, maintainable by local operators)

3.3 Connectivity Modes

  • Fully offline (default)
  • Local mesh only
  • Optional gateways for NGO-supervised updates

No dependency on commercial cloud services. This ensures resilience in conflict zones, emergencies, low‑bandwidth areas, and politically sensitive contexts.

4. Child Protection: Built‑In, Not Added On

4.1 Zero permanent child identity

No accounts, tracking, profiles, analytics, or behavioral logging. A child’s presence is physical, not digital.

4.2 Resettable within 60 seconds

A safeguard against unauthorized access, harmful data accumulation, and misuse by authorities.

4.3 No data leaves the room

Aligns with:
- UNICEF “AI for Children” Principle 1 (Protect Children’s Data)
- Save the Children “Safeguarding in the Digital Age”
- UNHCR Data Protection Handbook

4.4 Transparent by design

Every configuration, action, and permission is inspectable. This supports trust from families, community oversight, and local accountability.

4.5 Harm‑prevention architecture

The system cannot be repurposed for surveillance, profiling, or behavioral manipulation. This is what we call structural protection:

“Protect children not with rules, but with architecture that cannot be misused.”

5. Pedagogy: Low Tech, High Agency

5.1 Hands-on learning

Children learn through programming, creating files, making lights blink, understanding networks, and exploring cause → effect.

5.2 No extractive metrics

No performance dashboards, behavioral scoring, or engagement manipulation. Learning remains child-led, curiosity-driven, unmonitored, and unscored.

5.3 Local AI as a teaching assistant, not a replacement

AI runs offline. It answers questions, supports exploration, and explains concepts. It never stores child identity and has no remote backend.

6. Deployment for Field Operations

6.1 Suitable environments

  • Refugee camps
  • Emergency classrooms
  • Child-friendly spaces
  • Rural low-tech schools
  • Community learning centers
  • Mobile education units

6.2 Infrastructure requirements

  • Power (grid, solar, battery)
  • Enclosed physical space
  • Basic local network via router or hotspot
  • NGO staff with minimal technical familiarity

6.3 NGO benefits

  • Zero vendor lock-in
  • No recurring licensing costs
  • Full offline capability
  • Simplified troubleshooting
  • Community ownership
  • Full transparency for audits

7. Humanitarian Alignment

7.1 United Nations Principles

  • ✔ CRC – safety, privacy, participation
  • ✔ UNHCR – child protection standards
  • ✔ UNESCO – digital inclusion
  • ✔ UNICEF – AI for Children principles

7.2 Save the Children

  • ✔ Digital safeguarding
  • ✔ Do-No-Harm digital interventions
  • ✔ Protecting identity in crisis
  • ✔ Evidence-based child learning support

7.3 INEE Minimum Standards

  • ✔ Community participation
  • ✔ Safe learning environments
  • ✔ Relevant, contextual learning content
  • ✔ Feasible infrastructure

8. Policy Recommendations

For UNICEF & Save the Children programs:
- Adopt local-first infrastructures in EiE and CP contexts
- Avoid cloud-reliant ed-tech in high-risk regions
- Use child-safe architectures that prevent misuse by design
- Enable communities to host and maintain their own systems
- Treat data minimization as a core safeguarding requirement
- Support open-source, audit-friendly digital tools

9. Conclusion

The Crumbforest provides a future-proof, rights-centered, low-cost, open-source model for digital learning. It offers NGOs a way to deliver safe digital education without risking children’s data, without corporate dependency, without surveillance, and without internet requirements.

It is a structural commitment to children’s rights, encoded not in policy documents alone, but in architecture, source code, permissions, network design, and the absence of risky features.

The Crumbforest is not a platform. It is a promise made operational.

A promise that says:

Every child deserves to learn safely — even offline, even in crisis, even without identity, even without the cloud.

🌍💙
For children. Everywhere. Always.