StudyRoom Privacy Policy

Version: 2026-08-20.2 Applies to: studyroom.cc and the StudyRoom application (web app, all subjects and features)

StudyRoom is operated by PIAZR Pty Ltd (ABN 69 660 867 390), 21A Eungai Pl, North Narrabeen NSW 2101. "StudyRoom Academy" is a registered business name of PIAZR Pty Ltd; "StudyRoom" is the product name used throughout this site and application.

For students: the short version

StudyRoom keeps track of what you've studied and how you're doing so it can give you better help — things like which topics you've mastered, your practice answers, and what you talk about with the AI tutor. A teacher or parent linked to your account can see a summary of your progress, and sometimes StudyRoom's AI notices you might need a hand and lets them know. Some of what you type or say is sent to outside AI companies (like Google, Anthropic, OpenAI, or DeepSeek) so they can generate a response — that's how the tutor works. If you're under 15, your parent or guardian has to say yes before you can upload photos of your work, use voice practice, or send English Writing to those overseas AI companies for scoring. You can ask a parent, teacher, or admin what StudyRoom knows about you at any time.

1. Who we are and what this document covers

StudyRoom ("we", "us", "the platform") is an AI-assisted tutoring product for school students in Australia, used directly by students and by their parents/guardians and teachers. This policy explains what personal information we collect, why, how it's used (including by AI systems), who it's shared with, how long we keep it, and what rights you have over it.

Quick summary: this policy applies to everyone with a StudyRoom account — students, parents, teachers, and admins.

2. What we collect, and why

Category	Examples	Why we collect it
Account/profile	name, email, password (hashed), role, year level, state, home language, timezone	To create and run your account, and to tailor content to your year level and curriculum (AU state)
Date of birth	date of birth	Required at registration so we can apply the right privacy protections if you're under 15 (see §6). Currently only collected at credentials registration — accounts created via Google sign-in do not yet have a DOB on file, which is a known gap listed in §15
Learning activity	chat messages with the AI tutor, practice/exam answers, writing submissions, mastery/progress data, misconceptions, study plans, streaks	This is the core of how the tutor and planner work — without it we can't personalise anything
Photos of handwritten work	photos submitted for AI marking/feedback	Lets the AI read and mark handwritten maths/exam work. Gated by parental consent for under-15 students (§6). Raw photos are not stored after marking
Voice / debating practice	live spoken conversation with an AI debate partner, saved as a text transcript	Powers the live voice debate-practice feature. Gated by parental consent for under-15 students (§6). Audio itself is never stored — only the transcript, and only until the retention window closes (§8)
AI usage metadata	which AI model handled a request, token counts, cost, timestamps — not the message content itself	Lets us monitor and cap AI spend, and audit which model produced a response
Guardian/teacher links	guardianship requests and approvals, class rosters	So parents can see their own child's progress and teachers can see their own class, and nobody else's
Consent records	photo-upload consent, voice-practice consent, English Writing overseas-AI consent (`ai_writing_overseas`), terms/privacy acceptance	To keep an accurate, auditable record of what was agreed to and by whom
Marketing attribution	first-touch campaign/landing cookies (`tf_attribution`, `tf_landing_assignment`)	Set only on public `/go` links. Materialised at registration for a parent account only — never attached to a student's learning activity. See the Cookie Notice

Quick summary: we collect what's needed to run a tutoring product and keep it safe for kids — not more.

3. How the AI works with your data

When you chat with the AI tutor, submit writing, ask for a photo to be marked, or use voice debate practice, the relevant content is sent to a third-party AI provider to generate a response. We route text requests through OpenRouter (a US-based AI infrastructure company), which in turn sends the request to whichever model is active — this may be DeepSeek (a Chinese AI company), Google (Gemini), Anthropic (Claude), or OpenAI (GPT), depending on what an admin has configured.

Vision (photo marking) always uses a Google Gemini model.
Live voice debate practice uses Google's Gemini Live API directly (not via OpenRouter).
Text-to-speech (read-aloud) is generated on our own server (not sent to a third party) and cached.
Automated progress summaries and nudges (see §4) use Anthropic's Claude Haiku model.

Quick summary: your messages may leave Australia and be processed by AI companies based in the US or China, because that's how the tutor generates its responses. We don't sell this data or use it for advertising.

We keep a record of which model handled each request and its cost, but we do not store the actual prompt/response content in that usage log — it's metadata only. Separately, if we ever turn on optional debugging/tracing infrastructure (Langfuse) with content-tracing enabled, full prompt/response text could be logged for troubleshooting — as of this document's version, we have not confirmed whether this is active in production; this is flagged in the review checklist at the end of this document.

4. Automated decisions the AI makes about your progress

For mathematics (currently the only subject this covers), StudyRoom runs an automated process every few hours that looks at a student's recent activity — things like how many practice reviews are overdue, unresolved misconceptions, and days since last activity — and decides whether to:

do nothing,
send the student a gentle, encouraging nudge, or
escalate: post a note to the student's teacher and linked parent(s), flagging that the student may need attention.

This is genuinely automated — no person reviews it before it's sent. The message itself is usually written by an AI model (Claude Haiku), grounded only in the real signals described above; if that AI call fails, a plain template message is used instead.

How to see and contest this: every automated nudge or escalation is recorded and visible to parents and teachers on the student's "AI Report" page, where a parent or teacher can reply, ask the AI a grounded follow-up question, or mark an escalation resolved. Students do not see this feed directly. A teacher can also mute the automated process for an individual student at any time.

Quick summary: an automated system — not a person — decides when to flag a student's progress to their teacher/parent. It's not a clinical or professional judgement, just a pattern in the platform's own data, and a teacher or parent can always follow up, question it, or turn it off for that student.

How you're told about a nudge or escalation: these are shown in-app on the "AI Report" page and can also be delivered by email. A parent or teacher (never a student) who has opted in may also receive supported alerts over WhatsApp — see §7 for how that works, who provides it, and how to opt out. WhatsApp is optional and is only sent after phone verification and the extra overseas-disclosure consent in §9. What this does not do (yet): it does not currently look at English/writing activity, and does not detect real-time distress during a live voice session — see §5's live-voice section for that specific gap.

5. Live voice practice — extra detail

If a student uses live voice debate practice: their microphone audio is streamed to Google's Gemini Live API in real time to hold a spoken conversation; the audio itself is never saved — only a text transcript of the session. Sessions are capped (currently 15 minutes each, 30 minutes/day per student) and time-boxed. The AI is instructed to stop and suggest talking to a trusted adult if a student expresses distress, but this is a prompt-level instruction, not an independently verified real-time safety check — after the session ends, the transcript is automatically scanned for safety concerns, and if anything is flagged, the student's teacher(s) are notified.

Quick summary: voice practice conversations aren't recorded as audio, only as text, and there's an automatic after-the-fact safety check on top of the AI's own in-conversation instructions.

6. Children's data and consent

We treat any student whose date of birth is unknown, or who is under 15, as needing parental/guardian consent for three specific capabilities:

Photo uploads (for AI marking of handwritten work),
Voice practice (live spoken debate sessions), and
English Writing overseas AI scoring — sending writing submissions to the overseas AI providers in §3 (ai_writing_overseas consent).

A parent/guardian must be linked to the student's account (via our guardianship-verification process) before consent can be granted for an under-15 student; a student aged 15 or over can consent for themselves. If we don't know a student's date of birth, we treat them as under 15 by default (fail-closed) rather than assuming they're old enough.

For general use of the tutor (chat, practice, exams) below the age of 15, we rely on the account having been set up and approved by a parent, teacher, or admin in the first place — every new account requires admin approval before it can be used at all. English Writing that would be scored by an overseas model is the extra gate above, not covered by account approval alone.

Guardian identity verification today is an admin's own judgement call (e.g. "known family," confirmed by the student), not an independent identity check — we're flagging this honestly rather than overstating how rigorous it is.

Quick summary: photo uploads, voice practice, and English Writing scored by overseas AI all need a parent's yes for under-15 students. If we don't know a student's birthday, we treat them as needing that yes.

7. Who we share information with

AI providers described in §3, to generate tutoring responses.
Linked parents/guardians and teachers, who can see a student's progress summary, writing feedback, and the AI Report feed described in §4 — never a student's raw private chat conversation with the tutor.
Email delivery via Resend, for notifications you've opted into. Commercial electronic messages identify us and include a working unsubscribe facility. We action unsubscribe requests promptly (ACMA guidance: generally within 5 working days). Factual messages such as a one-time verification code are not marketing.
Opt-in WhatsApp notifications to parents and teachers only — never students. If a parent or teacher chooses to, they may receive supported alerts (such as progress digests and status alerts) over WhatsApp, sent and delivered directly via WhatsApp (Meta Platforms) — Meta's Cloud API, with no third-party messaging intermediary. Recipients verify their phone number and explicitly enable WhatsApp (with a separate cross-border consent, see §9) before any WhatsApp message is sent, and can reply STOP at any time to turn all WhatsApp messages off (reply START to opt back in where eligible; email and in-app notifications continue). Only approved, factual message templates are used — never free-form or AI-composed text.
WhatsApp (not SMS) is used to deliver a one-time code when verifying a phone number for the WhatsApp feature above, via a Meta authentication-message template.
We do not sell personal information, and we do not use student data for advertising.

8. How long we keep information

Honestly: most data categories do not yet have a formal retention/deletion schedule — this is a known gap listed in §15, not something we're hiding.

The one exception today is debate transcripts: these are automatically deleted after 30 days by default (an admin-configurable setting), via a daily automated job.

Everything else — chat history, practice/exam records, writing submissions, notifications, AI usage logs — is currently retained indefinitely while your account is active. We will update this section with a full schedule once one exists.

Quick summary: voice-practice transcripts auto-delete after 30 days. Everything else is kept indefinitely for now — we're working on a proper retention schedule.

9. Cross-border data

Some of the providers we use are based outside Australia, so personal information may be processed on servers outside Australia:

AI providers — the United States for OpenRouter, Google, Anthropic, and OpenAI; China for DeepSeek — process the content you share with the tutor (§3).
Meta/WhatsApp delivers the opt-in WhatsApp notifications described in §7; processing may occur outside Australia.

Accountability (Australian Privacy Principle 8.2). PIAZR Pty Ltd remains accountable for personal information disclosed to these overseas providers. We take reasonable steps to ensure they handle it in a way consistent with the Australian Privacy Principles. Before any WhatsApp notifications are enabled for a parent or teacher, we capture a separate, one-time cross-border consent covering this overseas disclosure.

10. Your rights

Access & export: you (or your guardian, if you're a student) can export your own writing data (submissions, feedback, growth history, goals) at any time.
Deletion: you can request deletion of your writing data specifically, the same way. This does not currently delete your whole account or other data types (chat history, practice records, exam attempts) — for a full account deletion, contact an admin directly; there is no self-service option for that yet.
Correction: most profile fields can be edited directly in Settings/Profile. Date of birth cannot be self-edited (to prevent a student ageing themselves past the under-15 protections above) — ask an admin to correct it if it's wrong.
Questions or complaints: email our privacy contact at privacy@studyroom.cc, or contact your school/platform admin. Please include enough detail for us to identify the account and the issue. We will acknowledge and respond to privacy complaints.

If we do not respond within 30 days, or you are not satisfied with our response, you can lodge a written privacy complaint with the Office of the Australian Information Commissioner (OAIC). The OAIC asks you to complain to us first. See Lodge a privacy complaint with us (OAIC page updated 2 March 2026; retrieved 20 August 2026). The OAIC does not take privacy complaints over the phone. Online form or email: oaicintake@oaic.gov.au; post: GPO Box 5288, Sydney NSW 2001. You may also call the OAIC Enquiries Line on 1300 363 992 for help lodging a complaint.

PIAZR Pty Ltd ABN 69 660 867 390 is the organisation to name if you complain to the OAIC about StudyRoom.

11. Cookies and similar technologies

StudyRoom sets a small number of first-party cookies so you can stay signed in, so English Writing can remember a year-band preference, and so public /go landing links can attribute a later parent registration. Attribution cookies are never attached to a student's learning activity. We do not currently run a third-party advertising pixel or a cookie-consent banner. Australian law does not currently require a GDPR-style cookie wall for this set of first-party cookies; this notice and the table in the Cookie Notice are our APP 5 collection notice for that information.

The names, purposes, and life of each cookie are in the Cookie Notice.

12. Notifiable data breaches

If we have reasonable grounds to believe an eligible data breach has occurred under the Notifiable Data Breaches scheme, we will promptly notify individuals at risk of serious harm and notify the OAIC. An eligible data breach is unauthorised access to, or disclosure of, personal information we hold (or loss of it in circumstances where unauthorised access or disclosure is likely) that is likely to result in serious harm, where we have not been able to prevent that likely risk with remedial action.

A notification will include our name and contact details, a description of the breach, the kinds of information involved, and recommended steps for affected people. See the OAIC page Report a data breach (retrieved 20 August 2026).

We do not currently publish a standalone incident-response runbook on this site. That operational gap is listed in §15.

13. Children's Online Privacy Code (status)

The Australian Privacy Principles already apply to how we handle children's personal information. Separately, the Privacy and Other Legislation Amendment Act 2024 requires the OAIC to develop a Children's Online Privacy Code — an APP code for certain online services likely to be accessed by children.

As of 20 August 2026 that Code is not yet registered and is not in force. The OAIC's Code page (published 19 March 2026, updated 5 August 2026) states the Code must be finalised and registered by 10 December 2026. Public consultation on an exposure draft ran 31 March 2026 to 5 June 2026. We will review this policy when the Code is registered. We do not claim the Code applies to StudyRoom today.

Source: Children's Online Privacy Code (OAIC; retrieved 20 August 2026).

Our under-15 / unknown-DOB fail-closed gates in §6 are a product rule we already operate under the APPs. They are not a claim that we already comply with a Code that is not yet in force.

14. Changes to this policy

We'll update the version number and date at the top of this document whenever we make a material change, and you'll be asked to re-accept it the next time you sign in if the version has changed since you last agreed.

15. Residual gaps we are not hiding

These are true today. We list them so this policy does not overclaim:

No full retention/deletion schedule except debate transcripts (§8).
Google sign-in accounts may still lack a date of birth; those accounts are treated as under 15 until a DOB is on file (§2, §6).
We have not confirmed whether optional Langfuse content-tracing of prompts/responses is on in production (§3).
Self-service export and deletion cover writing data only; full account deletion is admin-assisted (§10).
There is no in-app cookie dashboard (§11 and the Cookie Notice).
Guardian identity checks are an admin judgement, not an independent ID verification (§6).
No published incident-response runbook on this site (§12).

Dated regulator pages supporting the APP, Spam Act, and Children's Online Privacy Code status statements were retrieved on 20 August 2026 (OAIC, ACMA, ACCC). We will update this policy when those instruments change, including when that Code is registered.