Healthcare

Children's psychosocial health, with hospitals.

We are working with hospitals on breakthrough innovations in psychosocial healthcare for children: bringing what reaches a child online into their care.

Guardii runs on the child's own phone. Nobody reads their messages. A clinician sees what the child and family choose to share.

Startmatebacked
Official Meta Business Partner
The gap

More and more of the harm starts in a child's private messages.

Anxiety, self-harm, eating concerns and disclosures of abuse often trace back to what a child is living through online. Clinicians rarely see it: children don't say, parents don't know, and by the time it surfaces the trail is gone. No assessment has been able to reach it, until now.

How it works

The phone notices the pattern.The family decides what care sees.

Guardii looks for the shape of harm (targeting, escalation, coercion) on the device. When it finds one, the person it protects knows first, and a person decides what happens next. Nothing goes further without consent.

01

On the device

Where detection runs

On-device
No server-side capture

Connectors and detection run on the phone. Messages stay there.

02

Patterns

Not transcripts

Over time
Relationships, not keywords

Targeting, escalation and coercion are read across a relationship, never from a single message.

03

Severity and confidence

Kept apart

Pinned
Model, rules, thresholds

Every result carries how serious it is and how sure we are, separately. No naked scores.

04

The person knows first

Then their guardiian

First
The protected person

The person being protected sees what is raised before, or as, anyone else does.

05

Human decision

People make the call

Consent
Before anything leaves

A person decides what happens next. Evidence goes further only with consent, and police only after human review.

Patterns, not transcripts. People, not autopilot.

What reaches care

Online harm, brought into the assessment.

What is reaching the child

Grooming, sextortion, bullying and self-harm signals, described as patterns over time rather than a pile of screenshots.

Shared on the family's terms

Nothing reaches a clinician unless the child and family choose to share it. That consent travels with it.

Context for the assessment

It sits beside the clinician's own assessment and the conversation with the child. It informs care; it does not replace judgement.

Routed to the right team

Self-harm goes to care, exploitation to child protection, along routes the hospital sets with the family.

What we will not do
  • Give a clinician, a hospital or anyone else a window into a child's messages.
  • Diagnose, or hand anyone a risk score in place of a clinician's judgement.
  • Share anything without the child and family knowing.
  • Treat a child asking for help as evidence against them.
Research

Built with clinicians, in paediatric child protection.

Our research with hospital partners brings these patterns into paediatric child-protection assessments, in English and Arabic, reviewed by native speakers. We start with a limited-scope research project ahead of any wider rollout.

Read the research
Briefings

Bring online harm into children's care.

We brief hospitals, paediatric teams and child-protection services on the consent model, what reaches a clinician and how a research project starts.