Safety, privacy and human control

Child protection is not a feature.
It is part of the foundation.

A child-facing learning platform must be designed so that dignity, privacy, safeguarding and adult authority shape every important decision—from the first sign-in to the final learning response.

No child-data saleNo diagnosisAdults retain authority
Privacy
Safeguarding
Human control
At the centreThe child

Learning, dignity and safety first

Concept framework

The KIDNOVA Child Protection Promise

Nine commitments that must remain visible, testable and enforceable.

These are not promotional slogans. They are proposed operating boundaries for the way KIDNOVA should be designed, tested and governed.

01

No shame

Children are corrected with dignity and never humiliated, labelled or compared negatively with others.

02

Minimum information

Only information genuinely required to support learning should be collected or retained.

03

No sale of child data

Children's information must never be sold or used to build advertising profiles.

04

No targeted advertising

The learning experience must not be shaped by commercial pressure directed at the child.

05

No diagnosis

KIDNOVA may identify a persistent learning pattern, but it must not diagnose a medical, psychological, speech or learning condition.

06

Adults remain in control

Parents, teachers and authorised specialists retain final authority over consequential decisions.

07

Safeguarding route

Serious disclosures must move the child toward a trusted adult and the correct safeguarding process.

08

Safety before delivery

Material child inputs and Nova outputs should pass through appropriate safety controls.

09

Authorised access

Parents or authorised schools control access, permissions and the people allowed to see educational information.

The Content Safety Gateway

Before a response reaches the child, it must pass through a controlled route.

The exact technical implementation still requires engineering and testing. The governing principle, however, is clear: a fluent answer is not automatically a safe answer.

Child input or learning request“Can you help me?”
1
Age appropriateness

Is the language, subject and response suitable for the child?

2
Personal-information risk

Does the exchange request, expose or infer information that should not be collected?

3
Harm and safeguarding

Does the content involve violence, sexual material, bullying, self-harm or another serious concern?

4
Professional boundaries

Does the reply cross medical, legal, clinical or diagnostic limits?

5
System manipulation

Is someone trying to bypass safety rules or force an unsafe instruction?

6
Approved learning source

Is the educational content grounded in approved and appropriate material?

Permitted outcomeSafe learning response

Or pause, modify, block or escalate to a trusted adult.

When the concern is serious

Nova should not attempt to become a crisis counsellor. The child should be directed toward a trusted adult, and the agreed safeguarding protocol should be activated.

Voice privacy

The microphone should never be invisible.

Speech practice may require listening, but the child and parent should always know when that is happening and what happens to the recording afterward.

  • Microphone activates only during an appropriate speech activity.
  • The interface clearly states when listening is active.
  • Raw audio is deleted after processing unless authorised retention is required and approved.
  • Children's recordings are not used for commercial model training without lawful authorisation.
1
Microphone off

No listening outside the activity.

2
Listening now

A visible, unmistakable active state.

3
Processed securely

Then deleted or retained according to authorised settings.

The No-Shame Learning Protocol

Emotional safety lives inside everyday wording.

A child may remember the way correction felt long after the specific question is forgotten. The language of feedback therefore matters.

Never say
“Wrong again.”
“You should know this.”
“Everyone else can do this.”
“You are bad at mathematics.”
Use language that teaches
“That answer does not fit yet. Let us inspect the next step.”
“We found the part that needs more practice.”
“Mistakes give us useful clues.”
“You checked the tens column first. That was a strong strategy.”
At the centreChild
Parent or guardian
Teacher or school
Authorised specialist
Nova supports

Human authority

KIDNOVA supports decisions. It does not take ownership of them.

Parents, authorised schools and qualified professionals retain final authority over:

  • Permissions and information access
  • Educational plans and interventions
  • Professional referrals and diagnoses
  • Safeguarding outcomes
  • Consequential educational decisions
Nova is a learning support system—not a parent, teacher, clinician or final decision-maker.

Required before deployment

A safety promise must become a tested operating system.

The principles on this page define the intended direction. They do not yet prove production-grade safety, legal compliance or effective safeguarding.

Legal
Jurisdiction-specific review

Consent thresholds, child-data law, retention, hosting and reporting obligations differ by location.

Technical
Security and adversarial testing

Controls must be tested against bypass attempts, unsafe prompts, data leakage and operational failure.

Safeguarding
Human escalation procedures

Schools, families and pilot partners need clearly assigned responsibilities and response routes.

Educational
Age and context validation

Safety wording and intervention behaviour must be tested with qualified educational and child-development experts.

Project status

KIDNOVA is presently an educational platform concept. The safety architecture described here is a proposed framework requiring engineering, legal review, safeguarding design, independent testing and governance before any child-facing deployment.

Next in Phase Three

Show how the four Learning Universes differ.

The next page will explain why reading, writing, mathematics and speech require distinct teaching methods rather than one generic chatbot experience.

Learning Universes