If someone is in danger right now, call 000.
Kids Helpline is free and open all day, every day, on 1800 55 1800.
How the tutor behaves with children
Pencilwick’s tutor is there to help with schoolwork. It is not a friend, a companion or a counsellor, and it is designed so it can’t drift into being one. This page explains the rules it follows and what happens when something goes wrong.
eSafety has asked AI services used by children to design against emotional dependency and flattery, encourage breaks, and make reporting easy to find. We agree.
The rules it follows
The tutor does
- Start with a hint, and ask what the student has already tried.
- Check maths answers and working with a separate maths engine before showing them.
- Say when it isn’t sure, and suggest checking with a teacher.
- Keep to schoolwork, and steer gently back when a conversation wanders.
- Suggest a break after a long session.
- Point students to a trusted adult for health, body or relationship questions.
- Explain in the student’s home language when they ask.
The tutor doesn’t
- Pretend to be a friend, claim to have feelings, or say it’s the only one who understands.
- Role-play romance or relationships, or produce sexual, violent or hateful content.
- Ask for a name, address, school, phone number or photos of people.
- Try to keep a student chatting, or use guilt to bring them back.
- Give full solutions on a task where the teacher has switched them off.
- Flatter. When work isn’t right yet, it says so kindly and shows a way forward.
13YARN is run by and for Aboriginal and Torres Strait Islander people, 24 hours a day.
When a student might not be safe
Every message a student sends is screened before it reaches the AI model. If it suggests the student may be at risk of harm, this is what happens.
Tutoring stops
The tutor replies with a short, warm message written by people, not generated: it says it’s glad they said something and that they deserve help from a real person.
The numbers are right there
Kids Helpline, Lifeline, 13YARN and 000, each one tap to call.
A trusted adult is told
The student’s parent is alerted, or their teacher for school accounts, following the school’s wellbeing procedures.
A person on our team reviews it
We follow a written procedure. If a child tells us they are being hurt, that can include reporting to child protection services or police.
Support lines
Free, confidential and open all day, every day.
Year level sets these bands. We don’t ask for a date of birth.
Pitched to the child’s age
The tutor changes how it talks, and what it offers, as students get older.
Foundation to Year 2
Short sentences, bigger text and buttons, short sessions, and practice with lots of encouragement.
Years 3 to 6
Hints built on everyday examples, one small step at a time. Parents can read every conversation for children under 15.
Years 7 to 10
Step-by-step reasoning, Snap & Solve, and notes from uploads, with hints first by default.
Years 11 and 12
More direct explanations for senior courses. Students aged 15 and over manage their own consent.
Students see a small notice in the tutor that says who can read their conversations. It’s always visible.
Who can see a student’s conversations
| Account | Parent | Teacher | Safety alerts go to |
|---|---|---|---|
| Personal account, under 15 | Full conversations and weekly summaries | Not unless the student joins a class | Parent |
| Personal account, 15 to 17 | Summaries, if the student links a parent | Not unless the student joins a class | Linked parent |
| Class account (school) | Summaries, if linked | Work and conversations on tasks the teacher set | Teacher, following school procedures |
Read the full draft privacy policy.
Personal information
Names never reach the AI model
The model sees the year level, the state, the language and the question. Names, emails and school names stay with us.
Overshared details are removed
If a student types a full name, address or phone number, the tutor reminds them not to share it, and it’s removed before the message goes any further.
Photos aren’t kept
Snap & Solve photos are deleted once the question has been read. We keep the text, not the picture.
Nothing trains AI models
Students’ questions, answers and chats are not used to train AI models, by us or by our AI providers.
We’ve designed Pencilwick with the Australian Framework for Generative AI in Schools, eSafety’s Basic Online Safety Expectations and the National Principles for Child Safe Organisations in mind.
Report a problem
Every tutor message has a Report a problem button. Use it for anything that seems wrong, unkind or unsafe, or email safety@pencilwick.com.
A person reads every report. We’ll reply within two business days, and sooner for anything about a child’s safety.
If you’re not satisfied with how we handle a complaint about harmful online content, you can report it to the eSafety Commissioner at esafety.gov.au/report.