Blog

Every piece of software a school installs creates a record of a student somewhere. Usually it is a name and an email address — and in most Australian schools that email encodes the student's full name and year of enrolment. billy.smith@education.example is not anonymous data. It is a directly identifying record sitting in a vendor's database.

Dictation software has historically been no different. Every user needed an account, and for a school that meant handing over a class list.

That has changed, and it is worth understanding why it matters for procurement rather than just for marketing.

What should an Australian school ask a dictation vendor about student data?

Five questions, in the order that actually matters:

1. What identifying information do you require to activate a licence? If the answer is a student email address, the school is creating a vendor-held record of every student who uses an assistive technology.
2. Is audio recorded or retained? Live dictation and ambient recording are architecturally different things. Ask which one you are buying.
3. Is our content used to train your models? A growing number of tools reserve this right, and a meaningful share do it on an opt-out basis — which in a school means it happens by default to children nobody asked.
4. Where does processing happen, and what happens under failover? Most cloud services route to the nearest region and overflow elsewhere under load. A vendor claiming guaranteed in-country processing for a globally routed service is worth a follow-up question.
5. Where do transcripts live? On the vendor's servers, or on the student's own machine?

Why does anonymous serial licensing matter for student privacy?

Because the safest way to protect a record is not to create it. Under serial-number licensing, a school registers one administrator account — typically a generic IT or learning-support mailbox — and receives a block of serial numbers. A student enters a serial and starts dictating. There is no student sign-up, no confirmation email and no per-student profile.

If the school wants to know which student holds which serial, it keeps that list itself, in its own systems, under its own controls. The vendor never receives it.

That is a materially different posture from a privacy policy promising to look after data the vendor is holding. It removes the record rather than managing it, which means there is nothing to breach, nothing to subpoena and nothing to migrate when the contract ends.

Which dictation products support anonymous activation?

Speech Recognition Cloud supports serial-number licensing for schools and universities, with no student name or email address required. It also completed the South Australian Department for Education's security and privacy assessment.

Audio is transcribed and then purged — no session recording is created — and dictation history is held in a local cache on the student's own Windows profile rather than on a server. Dictated audio and text are not used to train or improve any model.

It is verbatim dictation rather than generative AI, which also shortens the academic integrity conversation considerably: it types what the student said, and does not compose, rewrite or summarise on their behalf.

The full technical detail — what is collected, what is not, where processing happens under failover, and what a school still has to manage itself — is set out in how anonymous serial licensing works for schools.

What are the limitations schools should know about?

Three, stated plainly.

It is Windows only. If your fleet is iPads or Chromebooks, this is not the product for you.

In-country processing is not contractually guaranteed. Processing routes to the nearest regional data centre and can overflow to the United States or Europe under load. Because the audio carries no student identifier, what crosses a border is unattributed speech rather than identifiable personal information — but if your framework requires fixed in-country processing regardless of identifiability, that is a genuine constraint.

Diagnostic and analytics data is collected for troubleshooting and product improvement. It does not contain dictated text or audio, but it is not nothing, and an assessor will ask.

How can a school trial this without disclosing anything?

The free Starter plan requires no account, no email address and no credit card. Twenty minutes of dictation per month is enough to test recognition on the voices and vocabulary that matter before any procurement conversation starts — and the trial itself creates no vendor record, for the student or the staff member running it.

If it works, the conversation starts from evidence rather than a demo. If it does not, nothing has been spent and nothing disclosed.

Voice Recognition Australia has been deploying speech recognition in Australian schools, hospitals, courts and government departments since 1999. If you would like help scoping a deployment or working through your own privacy impact assessment, get in touch.

Information accurate at time of publication. Verify current product details and pricing before purchasing.