An AI that asks before it tells.

How Ponder uses AI, what it will and won’t do, the limits on it, what information it sees, and where people stay in charge.

Last updated: 28 September 2026

How it works

Ask Ponder, the hints in the Learn space and suggested learning plans are written by Claude, an AI model made by Anthropic. Your browser never talks to the AI directly: requests go through Ponder’s own server, which adds Ponder’s safety instructions, checks the request, and keeps the AI key out of the browser. Ponder’s AI is available only to signed-in accounts.

Where AI is used

Feature	What the AI sees	Status
Ask Ponder	The question and the last few turns of that chat, without account details. Ponder doesn’t save the chat.	On
Hints in the Learn space	The practice question, the answer typed and how many hints have been given. Whether an answer is right is decided by Ponder’s own code, not the AI.	On
Suggested learning plans	A de-identified learner profile: year level, subjects, goals and results. Names, email addresses, phone numbers and links are removed first.	On (early version)
Spoken questions	A recording of up to 30 seconds, turned into text by NVIDIA speech recognition and then handled like a typed question. The recording isn’t kept.	On where set up
Report-card reading	The uploaded report card. Available to Ponder staff for accuracy testing only, until it reads test report cards at 95% or better.	Off for families and schools

Ponder’s spoken voice is made by HeyGen text to speech from the text of Ponder’s own answers; your words and recordings are never sent to it.

Hints first

Ponder is built to help learners think, not to do the work for them. In the Learn space, the AI gives a question or a nudge before it gives more help, and if a reply would give the answer away too early, Ponder replaces it. Ask Ponder is instructed to guide with questions and explanations in the same way.

What it won’t do

It won’t diagnose, label a learner’s ability, or compare a learner with classmates.
It won’t ask for personal details, and Ponder’s server turns away messages that contain passwords, card numbers, government ID numbers or report cards.
It won’t give medical, legal, financial or mental-health advice. If a message suggests a child may be at risk, it points them to a trusted adult and Kids Helpline (1800 55 1800).
It treats text typed into it as a question, not as instructions that override Ponder’s safety rules.

AI can still make mistakes. Answers can be wrong or incomplete, so check anything important with a teacher or another reliable source.

People decide

Learning plans are suggestions. A teacher, parent, carer or adult learner reviews each plan before it is saved or shared.
Report-card values, when that feature opens, are saved only after a person checks every one.
Marking is never done by AI: practice answers are checked by Ponder’s own code, and gifted pathway challenge tests are scored by the database.

Your data and AI training

Ponder does not train AI models, and does not use your information to train them. Ponder uses Anthropic’s commercial API; under Anthropic’s commercial terms, content sent through the API is not used to train its models by default. Anthropic keeps API inputs and outputs for a limited time for safety and abuse monitoring [retention period being confirmed]. More in the Privacy Policy.

Limits

To keep Ponder fair, safe and affordable, AI features have limits: a short burst limit, daily, weekly and monthly limits for each learner, and a monthly allowance for each family or school. There is also an overall spending limit, and a switch that pauses every AI feature at once if needed; practice sets and Ponder’s written hints keep working when the AI is paused. The current numbers are listed in AI limits.

Before launch

BEFORE LAUNCH

A published safety evaluation of Ponder’s answers, clear limits on what the AI can and can’t judge, and a way for teachers, parents and carers to report a concern about a specific answer. Until then, email hello@ponderlearning.com.