Trust and transparency — AIEdu
Last updated: 11 November 2026. This page is in English only.
Operator: Customer Experience Insight Pty Ltd trading as WA AI Hub (ABN 61 634 039 759)
Contact: our contact form
Current notices
No incidents to report as at 11 November 2026.
1. Our safety and security framework
We publish a Privacy Policy, Terms and Conditions, a Refund Policy, a Children’s Privacy Policy, a page on how we use AI and this page.
The protections in place include:
- Sign-in is handled by a dedicated identity provider. Passwords are stored hashed and we never see them.
- Request rate limiting. The limiter keeps a salted one-way hash of your IP address or ID, not the value itself.
- A content security policy and transport security headers.
- Same-origin checks on account actions.
- Scheduled deletion under the retention schedule in our Privacy Policy.
- You can download your data and delete your account yourself.
- Our AI call logs hold technical details and a one-way hash of the prompt, not the prompt text.
- Card details go only to the payment provider and never reach us.
Multi-factor authentication is not currently offered for accounts.
This is a description of how the service is built. It is not a certification.
2. AI models and independent evaluation
- Scoring and feedback. In Prompt Gym, Bias Detective and Output Critic, an AI model scores what you submit and writes feedback.
- Learning path and tutor. An AI model helps write your learning path text and gives short tutor hints.
- Translation. Pages shown in another language are machine-translated by an AI model. Legal pages are not translated.
The models come from OpenAI and Google. The full register of AI uses and automated decisions, and how to ask for a human review, is on the how we use AI page.
No independent third party has evaluated this product or the way it uses AI. There has been no outside audit, certification or penetration test. We will update this page if that changes.
The companies that build the models we use publish their own safety evaluations. Those are their evaluations of their models, not an evaluation of this product. See OpenAI and Google.
Our own checks:
- Automated tests run in CI on every change.
- Tests of the tutor’s refusal rule and of the sign-in requirement on AI features.
- A check that changes to AI and sign-in code carry a written change record.
These are run by us and are not independent.
3. Who can access your data
You and your organisation. Your account, progress, scores and certificates are tied to your sign-in. On a team plan, the team owner can see who was invited. Your pillar scores are shown in the team view only if you opt in, and your prompts and answers are never shown there. This is enforced by the application.
Us. One person, the owner of the business, can access production systems and data. There are no other staff or contractors with access.
Our tools. The owner uses AI-assisted operations tools to check that the service is running correctly. Those tools have read-only access to the database. Any change they suggest must be approved by the owner before it runs. Queries that could return personal information are restricted to AI models running on the owner’s own computer.
Our subcontractors. See section 4.
4. Our subcontractors and their boundaries
| Function | What it handles | Where |
|---|---|---|
| Hosting | Request data including your IP address | The application runs in Australia (Sydney). Delivery is global, and some logs may be in the United States |
| Database | Account and learning records (not your password) | Australia (Sydney) |
| Sign-in | Name, email, password hash or sign-in method, sessions, IP address and device details | United States |
| AI processing | Text you submit to AI features, and prompts built from your role and results | United States |
| Payments | Name, email, billing details and payment method | Global, including the United States |
| Contact form | Topic, name (optional), email and your message | United States |
Every subcontractor is engaged on its standard published terms. We have not negotiated separate agreements with them and we do not audit them; we rely on the commitments they publish. A named list is available on request.
For this product, each provider is also named in section 4 of our Privacy Policy.
The limits we apply ourselves:
- The application and the database are pinned to an Australian region (Sydney).
- We do not intentionally send your name or email address to AI providers.
- We do not keep the prompt text of AI calls in our logs.
- The providers’ published terms state that inputs and outputs sent through their APIs are not used to train their models.
5. How quickly we would tell you about an incident
If personal information we hold is lost, or accessed or disclosed without authorisation, we will begin assessing it within 72 hours of becoming aware of it. If the incident is likely to cause serious harm, we will notify the people affected and the Office of the Australian Information Commissioner as soon as practicable, and no later than 30 days after becoming aware. Where the GDPR applies, we will notify the relevant supervisory authority within 72 hours where that law requires it.
This service is run by one person. There is no round-the-clock on-call team, so an incident that happens overnight or while the owner is travelling may not be seen immediately.
6. How an incident would reach you
- Account holders: we would email the address on your account.
- People on the waitlist, and people who sent a message through the contact form: we would email the address you gave.
- Visitors without an account: we hold no contact details for you and cannot reach you directly.
- If we hold no contact details for you, look at the Current notices section at the top of this page. We will post a dated notice there.
To report a problem to us: use the contact form.
7. What we don’t publish, and why
We do not publish the prompts, scoring and ranking methods, pricing logic, abuse-detection rules or model files behind this product, or the findings of our internal reviews. Publishing them would make the service easier to copy or to attack. Where we describe a method publicly, the description is linked above.