W
WriteWiseCoach
← Back to home

AI Transparency Notice

Last updated: 7/30/2026

WriteWiseCoach uses AI to assess student writing against a fixed marking rubric and to generate written feedback. This page describes what the AI does, what it does not do, what data it sees, and how parents and teachers stay in control.

1. What the AI is used for

  • Rubric-based scoring.The AI reads the student’s draft and returns a numeric score per rubric criterion (for example: audience, cohesion, vocabulary, punctuation).
  • Written feedback. The AI writes short comments explaining each score, using verbatim quotes from the draft so the reasoning is checkable.
  • Coaching suggestions.On request, the AI produces concrete next-step suggestions in the student’s own voice (e.g. “Try replacing your third sentence with a rhetorical question”).

2. What the AI is NOT used for

  • It does notrewrite the student’s work in a way they could pass off as their own. Any suggestion the AI makes is clearly labelled as an example.
  • It does notreplace a teacher’s judgement. Teachers can moderate, adjust, or reject any AI-generated score before it appears on a formal report.
  • It is not a psychological or safeguarding assessment. Automated scanning of drafts for self-harm or harmful-content signals is on our roadmap but is not yet in production. When it ships, a SafetyEvent will be logged and the relevant teacher notified. Until then, do not rely on this service as a safeguarding tool.
  • It does not talk to the student conversationally, chat-bot style. All AI interactions are one-shot: submit a draft, receive one report. There is no open-ended chat surface.

3. Training and privacy

  • Student writing is never used to train or fine-tune AI models. Our contract with the AI provider prohibits training on customer data. See OpenAI’s Enterprise Privacy commitment for the current provider’s statement.
  • Names and identifiers are removed before sending. Our anonymisation pipeline strips out the student’s name, teacher names, school names, phone numbers, email addresses, and street addresses before any text leaves our servers.
  • All reports are generated on demand and stored in your account only. They are not shared with any third party beyond the AI call itself.
  • Data residency. Our database is hosted in Sydney (Australia, ap-southeast-2). The AI provider processes requests in the region documented in our Privacy Policy.

4. How the AI is calibrated

The AI is nota bespoke model — it is a general-purpose commercial model (see the current provider name in our Privacy Policy). We calibrate it for each rubric by including a small set of teacher-marked examplesin every request. This is called “few-shot prompting”. The examples never leave our servers persistently — they are sent with each request, used to anchor the AI’s scoring, and discarded when the request completes.

None of these examples are used to permanently train the AI. If we add or remove an example, the effect is immediate on the next request — not a training run.

5. Human oversight

  • A teacher can review, adjust, or reject any AI score before it appears on a formal report.
  • Parents can revoke their child’s AI-assessment consent at any time via the parent consent page. Once revoked, no further AI calls are made for that child.
  • Every AI-generated report includes the model name, the version of the prompt used, and the anchor examples that informed it, so the decision is auditable after the fact.

6. Known limitations

  • Any AI can make mistakes. Scores should be treated as a starting point for a conversation with a teacher, not as absolute truth.
  • The AI works best on the text types it has been calibrated for (persuasive, narrative, information-report, letter). Other genres may produce less reliable scores; teachers can flag such reports for review.
  • The AI does not judge factual accuracy of the student’s claims — only the writing craft. Historical, scientific, or personal claims in a draft are not verified.

7. Questions or complaints

If you have a question about how the AI reached a particular conclusion, or you believe a report is unfair, contact us at info@aiglobe.com.au. We will respond within one business day and can supply the raw AI request and response for that report on request from the account owner.