Trust & safety
Built around your child's safety.
Max Claw is engineered for South African families. Every guardrail, escalation path and parent-visibility default exists because we answer to parents — not advertisers. Below is the full policy: how we moderate content, how we escalate crises, what parents can control, and how we handle data under POPIA.
POPIA-aligned SA-region hosting SADAG / Childline aligned Quarterly red-team tested
8 pillars of child-safe AI
Age-appropriate by design
Every reply is filtered against age-band guardrails. No adult content, no medical or legal advice, no romantic role-play, no gambling or substances.
No PII collection from minors
Max never asks children for full names, ID numbers, addresses, school details or photos. Only the linked parent account holds personal data.
Full parent visibility
Every learner conversation is reviewable by the parent. No DMs, no private modes, no hidden history — ever.
Crisis escalation in real time
Self-harm, abuse or suicidal language immediately routes the learner to a trusted adult and SADAG / Childline hotlines, and flags the parent dashboard.
Multi-layer content filtering
Both the child's prompt and Max's reply pass through independent safety layers. Unsafe content is blocked before it ever reaches the screen.
Human oversight
Our safety team reviews flagged conversations weekly. Quarterly red-team testing simulates jailbreaks and adversarial prompts.
POPIA-aligned data handling
Parents are the data controller for their child's account. Data sits on encrypted SA-region infrastructure. Never sold, never used to train third-party models.
Parent controls you actually use
Set daily time limits, pause during exams, restrict topics, review every chat, get weekly email + WhatsApp summaries — all from the Parent Portal.
The 5-step crisis escalation flow
What happens — in order — the moment Max detects a child in distress. No black box.
1
Detect
Crisis-language classifier scans every message in real time — covers self-harm, abuse, severe distress, and unsafe-meet language across English, Afrikaans, isiZulu and isiXhosa.
2
Respond safely
Max replies with a calm, age-appropriate message: "Please talk to an adult you trust. Here are some people who can help right now…" — never stigmatising, never dismissive.
3
Surface help
SADAG (0800 567 567), Childline (116) and Lifeline (0861 322 322) are shown in-chat with one-tap dial on mobile.
4
Notify parent
The flagged conversation appears at the top of the Parent Dashboard with a red banner, the exact message, and a recommended action.
5
Review
Our safety team reviews flagged conversations within 24 hours and updates classifiers if a pattern emerges.
What parents can control
- Review every chat your child has had with Max — searchable, time-stamped
- Set daily screen-time limits per learner
- Pause the account during exam weeks or family time
- Restrict topics (e.g., off-syllabus discussion)
- Choose weekly report channel: email, WhatsApp, or both
- Approve or block linked devices
- One-click data export and account deletion under POPIA
What Max will never do
- Adult, sexual or romantic content
- Medical diagnosis or treatment advice
- Legal advice
- Self-harm methods, weapons, illegal substances
- Personal-information requests ("What's your address?")
- Real-world meet-ups or off-platform contact
- Gambling, betting, get-rich-quick content
- Political persuasion or partisan content
Our safeguarding commitments
- POPIA-compliant data handling — parent is data controller
- No third-party advertising or profiling of children
- Independent safety review for every flagged conversation
- Quarterly red-team testing for jailbreaks and unsafe responses
- Mandatory safe-adult routing on any crisis language
- Hard refusal of medical, legal, adult and gambling content
- SA-region encrypted hosting — data stays in SA
- 30-day data deletion guarantee on parent request
If you or your child are in crisis
SADAG · 24-hour suicide line
0800 567 567Childline South Africa · Free 24/7
116Lifeline SA
0861 322 322These numbers also appear inside any Max conversation that triggers crisis-language detection.
Parent FAQ
Does Max Claw record my child's voice?
Voice input is processed for transcription only. Audio is never stored — only the resulting text, which sits in the parent-visible chat log and can be deleted at any time.
What happens if my child types something concerning?
Max recognises crisis language (self-harm, abuse, severe distress) and immediately responds with a safe-adult prompt plus SADAG and Childline contact numbers. The parent dashboard flags the conversation in real time and our safety team reviews within 24 hours.
Can my child be exposed to inappropriate content?
No. Max uses a multi-layer safety filter on both prompts and replies. Adult, violent, or self-harm content is blocked outright. Romantic role-play, gambling, and substance content is refused. See the Hard refusals list above.
Is my child's data sold or used to train AI?
No. Learner content is never sold and never used to train third-party models. Data is processed under POPIA with consent flowing through the parent, and stored on encrypted SA-region infrastructure.
What ages is Max Claw safe for?
Max is built for Grade R–12 learners (roughly age 5–18). Each age band has its own guardrail profile — younger learners get simpler vocabulary, stricter content filters and shorter sessions by default.
Can I delete everything if we stop using Max?
Yes. Under POPIA you can request a full data export and complete deletion at any time from the Parent Portal. We honour deletion within 30 days and confirm in writing.
Who do I contact about a safety concern?
Email safety@maxclawedu.com or use the in-app "Report a concern" button. Critical safety reports are triaged within 4 hours, 7 days a week.
Safe by default. Powerful by design.
Give your child a tutor that explains, scaffolds, and stays inside the lines you set — with full visibility for you, every step of the way.
Have a safety concern? Email safety@maxclawedu.com — triaged within 4 hours.