KIARI Legal
Safeguarding Policy
KIARI is an AI tutor that talks directly with children. This policy sets out how we protect them, what happens if something concerning comes up in a lesson, and how to raise a concern with us.
Last updated: 22 September 2026
If a child is in immediate danger, do not use this page.
- Call 999. KIARI is not an emergency service and cannot respond in real time.
- Childline: 0800 1111, free, 24 hours, confidential, for any child.
- NSPCC Helpline: 0808 800 5000, for adults worried about a child.
- Samaritans: 116 123, free, 24 hours.
- To reach us urgently: 07841612909 or baileyturan05@gmail.com.
1. Scope and commitment
KIARI provides one-to-one AI tutoring to children, most of secondary school age. Although no human tutor is present and children never interact with each other, a child is still talking, out loud and unsupervised, to a system that listens. That creates a real responsibility, and we do not treat the absence of a human on our side as the absence of a safeguarding duty.
We are committed to the principle that the welfare of the child is paramount, and that safeguarding is everyone's responsibility.
This policy applies to everyone who operates or has access to KIARI. The designated safeguarding lead is Bailey Turan, who can be reached at baileyturan05@gmail.com or on 07841612909. KIARI is operated by one person, so the designated safeguarding lead is also the person who runs the service; we say so plainly rather than implying a larger team or a separation of roles that does not exist.
2. What KIARI is, and what it is not
Being precise about this matters, because over-claiming would be its own risk.
| KIARI is | KIARI is not |
|---|---|
| An automated tutor that detects a small number of serious signals in what a child says, responds to them in the moment, records them, and reports them to the account holder after the lesson. | A counselling, mental health, or crisis service. |
| A service that keeps a record of what was said, so a concern can be reviewed afterwards. | A monitored service with a human listening live. Nobody is watching a lesson as it happens. |
| A way a parent may find out their child is struggling with something. | A replacement for a school's designated safeguarding lead, social services, or the police. |
| Able to signpost a child to Childline and other support. | Able to intervene, contact anyone on the child's behalf in real time, or verify that a child is safe. |
3. What we monitor for
Every lesson is automatically checked for a small number of serious signals. We deliberately keep this list short, because a system that flags everything trains the people reading the flags to ignore them.
- Indications of self-harm or suicidal intent.
- Hate speech.
- Extremist content.
- Explicit or sexual language.
This monitoring is automated and runs on what the child says during a lesson. It is not a general surveillance of the child, and we do not use it to report ordinary bad behaviour, rudeness or frustration. A child having a bad day at maths is not a safeguarding event.
4. What happens when something is flagged
The same sequence runs every time.
| Step | What happens | When |
|---|---|---|
| 1. The child is responded to | The tutor answers the concern directly and seriously, in the same turn. It names the seriousness, asks the child to speak to their parent or guardian or another trusted adult, and gives them the Childline number so they can talk to somebody outside the family if they would rather. The tutor does not attempt to counsel, diagnose, or investigate, and does not brush past what was said. | Immediately, automatically. |
| 2. A record is made | The category of concern and a short snippet of the transcript that triggered it are saved, with a timestamp. | Immediately, automatically. |
| 3. The lesson carries on | The lesson is not stopped, paused, or locked. After the response above, the tutor hands the question back and the lesson continues normally. This is deliberate: a child who says something difficult should not have their lesson taken away from them for it, and an automatic stop would punish a misheard word as readily as a real disclosure. | In the same turn. |
| 4. The account holder is told | The parent or guardian receives the lesson report by email, which sets out the concern, what was said, and what the tutor did about it, alongside the rest of the lesson. | At the end of that lesson. |
| 5. External referral, if needed | Where there is reason to believe a child is at risk of significant harm, we will refer to the local authority children's social care team or the police, whether or not the parent agrees. | Without delay. |
There is no automatic human review step between the flag being raised and the parent being told. KIARI is run by one person and nobody monitors lessons as they happen, so a policy promising review before escalation would describe something that does not exist. The designated safeguarding lead reviews flags when they are raised with us, or when a parent gets in touch about one.
When we would act without telling the parent first
Normally the account holder is the right person to tell. There is one important exception: where telling them could put the child at greater risk, for example if the concern relates to somebody in the child's own household. In that situation we will seek advice from the NSPCC helpline or the local authority before contacting the household, and we will follow that advice.
We do not promise a child confidentiality that we cannot keep. If a child tells the tutor something indicating they are at risk, we will act on it.
5. Limits you should understand
We would rather state these plainly than let a parent assume a level of protection that does not exist.
- Detection is automated and imperfect. It will sometimes flag something harmless, and it may miss something real. It is not a safety net you should rely on.
- Nothing is monitored in real time by a person. A flag raised at 11pm is not read by a human at 11pm.
- You are told at the end of the lesson, not during it. The alert reaches you in that lesson's report, once the lesson finishes. It is not sent the instant a flag is raised, and it does not wait for anyone to review it first. If a lesson is interrupted and not resumed, the report is still sent.
- The lesson is never stopped automatically. The tutor responds to what was said and then carries on teaching. Nothing locks the account or requires you to unlock it.
- We cannot verify who is speaking. We know which account is in use, not who is in the room.
- We cannot see or hear anything outside a lesson.
6. How to raise a safeguarding concern with us
If you are worried about a child in connection with KIARI, whether or not the system flagged anything:
- Urgent: call 07841612909. If a child is in immediate danger, call 999 first.
- Non-urgent: email baileyturan05@gmail.comwith the child's account details and the date and rough time of the lesson, so we can find the right recording.
We will acknowledge a safeguarding concern within one working day and tell you what we have done. That is deliberately faster than the three working days the Reporting and Complaints Procedure commits to for ordinary complaints: a safeguarding concern is not an ordinary complaint and should not wait in the same queue.
General complaints that are not safeguarding matters go through the Reporting and Complaints Procedure.
7. Our own people
Access to lesson recordings and safety flags is restricted to those who need it to operate the service. At present that is one person, Bailey Turan, who is also the designated safeguarding lead. Bailey Turan holds a current DBS (Disclosure and Barring Service) check and has completed safeguarding training. No other person has access to recordings or safety flags, and access is not granted to anyone else without the same checks and training being in place first.
8. Records and review
Safeguarding records are kept securely, separately from ordinary lesson data, and are visible only to the account holder and to those operating the service. Retention: 24 months from the date the concern was raised, after which the record is deleted automatically. This matches the retention period published in the Privacy Notice and is the period actually enforced by the scheduled deletion job, so the promise and the system agree.
This policy should be reviewed at least annually, and whenever the monitoring or escalation process changes. Next review: 22 September 2027, or sooner if that process changes before then.