Vision

Our Vision

Children grow up with AI, whether we like it or not. Most AI systems are built to please and to hand out answers.

“We are building the opposite: a personal learning and development companion that protects children, lets them think for themselves and helps them create something.”

Not a therapist, not a replacement for parents, not an all-knowing advisor.

Three areas

Conversation, learning modules, workshops

SafeMentor has three areas. In all of them, the same mentor accompanies the child.

Today

Conversation and homework

The child asks, talks or gets stuck on a task. The mentor explains, asks questions and helps step by step, without doing the task for the child.

In development

Learning modules

Short, guided units on a topic, for example how AI works and when to trust it. Later also: understanding feelings, dealing with conflict, making good decisions. Each module consists of a few short units with questions to think about and a small task for everyday life.

Planned

Workshops

This is where something of their own is created. Over several weeks the child works on a project, and the mentor accompanies it in stages. In the writing workshop the child writes a story or a whole book, in the ideas workshop they develop their own business idea and try it out on a small scale. Every sentence and every idea comes from the child.

Principles

What we stand for

1

Thinking for themselves instead of collecting answers.

Children don't need an AI that answers everything, but one that helps them work it out themselves.

2

A good companion sometimes says no.

Honest and encouraging instead of flattering.

3

Safety is not a filter, it is the architecture.

It decides which data we collect, what we store and when people get involved.

4

What children create belongs to them.

Ideas, texts, projects and the decision about what to share.

5

A child's most important companions are people.

Our mentor leads towards them, never away from them.

Mentor

Our mentor's attitude

We don't look first at what a child is missing, but at what is already in them. Six sentences describe how the mentor treats every child, whether they are chatting, doing homework or working on a project of their own:

1

Connection

“I listen to you and take you seriously.”

2

Growth

“I discover your possibilities with you.”

3

Independence

“I help you find your own solutions.”

4

Challenge

“I trust you with the next step.”

5

Togetherness

“I support you in getting along well with others.”

6

Protection

“I look out for your safety and show you where to get help.”

At the end of a good conversation, a child should not think "The AI helped me" but "I did that myself". That is why, where it fits, a conversation ends with something the child can do, try or decide for themselves.

Inspiration

What we learn from the best

We are not reinventing the wheel. SafeMentor builds on four approaches that are among the best-researched in child and youth development. We are not affiliated with any of these organisations; their work inspires us.

1

Accompany: How good relationships work

The Search Institute describes five elements of relationships that help children grow: showing care, challenging growth, providing support, sharing power, and expanding possibilities.

What this means for us: Being nice is not enough. The mentor listens, encourages, disagrees when necessary, and lets the child help decide what kind of help they want.

  • Today

    A warm, honest tone instead of a yes-man AI; questions instead of ready-made solutions; praise for effort, not for speed.

  • Planned

    The mentor behaves the same everywhere, whether the child is chatting, doing homework or working through a learning module. The child helps decide how they are helped: “Do you want a hint, shall I ask you a question, or would you rather just tell me first?” And the mentor suggests projects and tasks for the real world.

2

Strengthen: A child's inner toolkit

CASEL organises social-emotional learning into five areas of competence: self-awareness, self-management, social awareness, relationship skills and responsible decision-making.

What this means for us: Children should not only get answers but tools: understanding feelings, solving conflicts, setting boundaries, making good decisions.

  • Today

    In case of conflict, bullying or worries, the mentor strengthens the child with proven response patterns, depending on whether the child is affected, is watching, or has crossed boundaries themselves.

  • In development

    Learning modules, first on AI literacy and critical thinking.

  • Planned

    Short learning modules on feelings, friendship and conflict, and decisions. Parents can unlock them; the child is invited, not obliged.

3

Persevere: Building resilience

The Center on the Developing Child at Harvard University shows that children who get through difficult times well almost always have at least one stable, reliable relationship with a supportive adult. In addition there are opportunities to build skills and to experience that you can make a difference yourself.

What this means for us: When a child fails, the mentor neither plays it down nor immediately supplies the solution. It helps make sense of what happened, discover their own strengths and find a next step. And it reminds them who can help in real life.

  • Today

    A help ladder for homework that helps only as much as needed; in case of worries, a pointer to trusted adults and the helpline.

  • Planned

    Workshops in which children create a project of their own in small stages, such as a book. Setbacks are part of the journey there, and in the end comes the question: what can I do better now?

4

Protect: Children's rights as a blueprint

UNICEF revised its guidance on AI and children in 2025 (version 3.0). It lists ten requirements for child-friendly AI, including safety from the start, strong data protection, transparency, fairness, accountability and genuine participation of children. New is a dedicated look at AI companions used by children.

What this means for us: Safety is not a filter in front of dangerous answers; it shapes the whole architecture.

  • Today

    Hosting and data in the EU; crisis texts written by humans, not generated by the AI; parents receive pointers in case of serious concerns, but no chat histories; chats are deleted after 30 days; children's chats do not train AI; the mentor says openly that it is an AI.

  • Planned

    Before a child starts something new, parents approve it, whether a learning module (for example on AI literacy) or a workshop. In workshops something of their own is created: in the writing workshop a story or a whole book, in the ideas workshop their own business idea, which the child tries out on a small scale. These results are kept in the child's personal project folder and are shared only if the child wants to.

Four building blocks

The synthesis

Each of the four approaches provides a building block:

InspirationSearch Institute
What it gives usthe quality of the relationship: how the mentor speaks and accompanies
InspirationCASEL
What it gives usthe development goals: what children learn and practise
InspirationHarvard Center on the Developing Child
What it gives usthe attitude in setbacks: strengthen instead of take over
InspirationUNICEF
What it gives usthe guardrails: children's rights as the blueprint of the technology

This is what SafeMentor grows out of: a personal learning and development companion for children aged 8 to 15. It is not a therapist, not a replacement for parents and not an all-knowing advisor. It is a patient sparring partner that asks questions, lets children practise, stays honest and points the way to the people a child really needs.

Overview

How it all fits together

One mentor, one attitude, several ways to learn

Child

App

Protection layer

detects dangers · hosted in the EU

Mentor

one attitude · for ages 8 to 15

Conversation

listening, explaining

Homework

help ladder instead of solution

Learning modules

AI, feelings, conflicts

Workshop: My Book

own projects

Approvals

Parent dashboard

approvals, pointers, no chat histories

Project folder

the child's work · only the child shares

in development or planned; everything else is in use today
How SafeMentor fits together · as of autumn 2026

Every message first passes through the protection layer. Behind it stands a mentor with one attitude, whether the child is chatting, doing homework, working through a learning module or working on a project of their own. Parents approve new areas and receive pointers, but no chat histories.

Honestly

What SafeMentor can do and what only people can

We want to be honest: an AI can do much of what good companions do. The most important thing it cannot do. Resilience research is clear on this: what carries children is a reliable relationship with a human being.

What SafeMentor can doBe available at any time, patient, without judging
What only people can doGive closeness: being there, a hug, a relationship over years
What SafeMentor can doAsk questions that make you think further
What only people can doBelieve in a child. Confidence works because a person visibly means it
What SafeMentor can doLet children practise decisions and think along
What only people can doHand the child real responsibility and stand by their decisions
What SafeMentor can doRecognise warning signs and show the way to help
What only people can doStep in when there is danger and protect a child
What SafeMentor can doDisagree honestly without shaming
What only people can doTruly know a child: voice, look, everyday life, history
What SafeMentor can doExplain things in a child-friendly way, support writing and learning
What only people can doComfort, advise, treat when it gets serious

And this too is part of being honest: an AI can be wrong. It can overlook a worry or raise a false alarm. That is why our crisis texts are written by humans, and why SafeMentor replaces neither supervision nor help from professionals.

Our conclusion: Over time the mentor makes itself unnecessary. It strengthens the child so that they can talk, argue, learn and try new things better with the people in their life.

Commitments

Our promises

No streaks, no tricks.

We don't measure ourselves by how long a child stays on the screen. After a while the mentor even encourages going outside.

Data stays in Europe.

Hosted in the EU, built in line with the GDPR.

Parents get pointers, not transcripts.

In case of serious concerns parents learn what it is about and how they can help, but they do not read along.

Children's chats do not train AI.

Honest about itself.

The mentor says that it is an AI and never pretends to be a human or a friend.

What a child creates belongs to them.

Go to the pilot phase