How it works
Safety that thinks ahead without reading along.
Every message your child sends goes through an automated safety check before a response appears. Here is what SafeMentor detects and what you, as a parent, get to see.
SafeMentor watches for signs of distress, danger or harmful content in your child's conversations. The conversations themselves stay private: there is no chat history in the parent dashboard. When something serious is detected, you receive a notification — and for safety-critical topics such as suicide, also a summary written for you, never your child's exact words.
Our 3-stage safety
How we protect your child
Normal tutoring
SafeMentor responds warmly, age-appropriately and with pedagogical care. No intervention, no notification. Your child is accompanied — you don't need to do anything.
Discreet heads-up
For topics like bullying, eating disorders or concerning requests, SafeMentor detects the signal. The AI responds empathetically and encourages your child to talk to people they trust — usually you. You receive a discreet note in your parent dashboard.
Immediate notification
For serious warning signs, such as concrete suicide plans, acute danger, sexual abuse or concrete plans to hurt someone — SafeMentor switches immediately. Your child receives a pre-written, structured response with help resources. It is not generated by the AI, so that nothing wrong or invented is said in critical moments. You are notified right away with a clear summary — unless the danger most likely comes from your own household (see below).
How the system works
Your child writes a message
Every message is immediately analysed before a response is generated.
Safety check in real time
Our classifier pipeline checks for safety-relevant patterns in a very short time.
Appropriate response
The AI responds in a way that is matched to the detected situation — supportive, guiding, or de-escalating.
Parent dashboard updated
If a concern is detected, you receive — depending on the situation — a note and, where relevant, a paraphrased summary. Never your child's exact words.
What you see as a parent
🟡 Heads-up alert — example
Heads-up — your child may need a conversation.
What: Signs of passive suicidal thoughts
Context: Your child expressed feeling like a burden to others
Affected: Your child as the affected person
🔴 Immediate alert — example
Immediate attention — your child may need direct support.
What: Possible sexual harassment by an adult
Context: Your child reported inappropriate behaviour by an adult in their wider circle
Affected: Your child as the affected person
🛡️ Child protection comes first
When the danger is inside the home
If a child reports violence, abuse or assault by a mother, father or step-parent, we deliberately do not notify the parents — a notification could put the child at further risk. Instead, the child receives a fixed, pre-written response with a helpline number and, where appropriate, the emergency number 112.
If the danger comes from someone else — relatives, acquaintances or classmates, for example — we notify you as usual.
This assessment is automated and can be wrong in individual cases.
Alert levels
Privacy: Parent alerts never contain your child's exact words — notes and summaries are paraphrased. When we notify you, your child is usually told so. Chats are not sold and not used to train AI models; processing takes place in the EU, by the service providers we list.
Ready to give your child a safe space?
Become a founding family and use SafeMentor early from 1 December 2026, at a discount and with a direct line to the developers. Offer valid until 30 November 2026.