ai.heart

AI Transparency Statement

A plain-language account of the AI you may talk to on this site: what it is, what it does and does not do, and how to raise a concern.

The Australian Government asks its own agencies to publish a statement in this shape. ai.heart is a private service, not a Commonwealth entity, and that requirement does not bind us. We are publishing this statement voluntarily, in the same form the Australian Government asks of its agencies, so that a government partner assessing this service can do so against a structure they already recognise. Nothing on this page should be read as a claim that this service is certified, endorsed, or officially assessed by any government body.

Why we use AI

This site runs a conversational AI. It lets a visitor ask a question by typing, or by speaking, and get a reply in return — in speech, in on-screen text, or both. We use it because it lets someone explore what ai.heart is and how it works through a conversation, rather than by reading a static page alone.

What it does and does not do

The AI is a conversational assistant on a web page. It can answer questions, hold a conversation, and describe ai.heart and its work.

It does not manage an account, process a payment, or take any action outside the conversation itself. It does not look up or hold information about who you are unless you tell it. It gives no professional advice — legal, medical, financial, or otherwise — and nothing it says should be relied on as such.

That you are talking to an AI

Every reply on this page is generated by artificial intelligence, not typed by a person in real time. This is stated here, and the page itself makes clear when the AI is speaking or listening. You can always use the service by typing — a microphone is never required.

Outputs are not human-reviewed

Each reply is generated by a large language model, run through a cloud AI provider (Microsoft Azure). No person reads or approves an individual reply before it reaches you. This is the classification the Australian Government's transparency standard asks agencies to state plainly where the public interacts directly with AI output, so we are stating it plainly here too.

That does not mean the reply is unchecked. Before it reaches you, and before what you say is passed to the model, both directions pass through this service's own screening layer. What that layer looks for and how it is built is not described here in detail, for the same reason a lock's mechanism is not published on the door — but its presence, and that it runs on every turn in both directions, is stated as fact.

What it will not decide about you

This AI makes no decision about any person. It does not determine eligibility for anything, does not make or influence a decision about a benefit, service, or application, and does not assess, score, or profile an individual. If you are looking for a decision about your own circumstances, this AI is not the channel for that, and nothing it says constitutes one.

How we screen what is said

Every turn of the conversation — what you say to it, and what it says back — passes through ai.heart's own screening layer before it is spoken or shown. This runs in both directions, on every turn, without exception. We describe it only at that level of generality here; the detail of how it is built is not published.

This is the same floor described on the live page itself. The two describe one mechanism from two angles: this statement for what it means for you, that page for what it looks like while you use it.

Where processing happens

Replies are generated by Microsoft Azure in Australia (Australia East) and screened on our own server in Australia. A few services sit outside Australia: speech for read-aloud (Cartesia, United States), web search (DuckDuckGo, United States) and our network provider (Cloudflare, global). In the model room, a signed guest may instead choose another model provider with their own key; that provider then generates the reply, and it is still screened by our governance system in both directions. Our separate privacy statement covers what is collected, how long it is kept, and your rights over it in more detail.

How we monitor it, and what we do when it gets something wrong

The screening layer described above is our first line of monitoring: it runs on every turn, in both directions, and is intended to catch and stop harmful or out-of-scope content before it reaches you.

We review what the service says and what our screening layer stopped on an ongoing basis during this pilot, and we change the system when a review shows it is needed. We have not yet fixed a published review cadence; when we do, it will be stated here.

If the AI gets something wrong — a factual error, a reply that is out of place, or behaviour that concerns you — we want to know. See the next section for how to tell us.

How to challenge or report something

If a reply from this AI was wrong, unhelpful, or concerning in any way, you can tell us. Email admin@aiheart.com.au and describe what happened — including, where you can, the approximate time and what was said. We treat every such report as something to look into, not something to dismiss.

A report sent to that address is read by the people who build and run this service. We have not yet published a guaranteed response time for this pilot; we aim to reply within a few business days.

Because this AI makes no decision about any person (see above), there is no determination of yours to formally appeal through this service. If you believe a reply caused you harm, the contact above is the way to raise it.

Our voluntary alignment

We are not bound by Australia's AI-specific regulation — no AI-specific guardrails have been legislated in Australia; a proposal for mandatory guardrails was superseded by the National AI Plan in December 2025, and none currently bind a private service such as this one. Where we refer to government frameworks below, it is only to say our approach is informed by them, never that we are certified against, or compliant with, any of them.

This service is informed by Australia's voluntary AI Ethics Principles:

It is also informed by Guardrail 6 of Australia's Voluntary AI Safety Standard:

"Inform end users about AI-enabled decisions, interactions with AI and AI-generated content." — Australia's Voluntary AI Safety Standard, Guardrail 6

This page — telling you plainly that you are talking to an AI, that its replies are not human-reviewed, and that it decides nothing about you — is how we act on that guardrail.

Who is accountable and how to contact us

Questions about this statement, or about how the AI on this site works, can be sent to admin@aiheart.com.au.

This service is operated by AIHEART PTY LTD (ACN 694 607 022), of Victoria, Australia, trading as AI HEART. That company is accountable for it, and is reachable at the address above.

This service is not currently part of a named government pilot. If that changes, it will be stated here.

This statement covers the conversational AI on this site. It does not cover how we collect or store personal information — see the privacy statement for that.

Last updated: 25 September 2026.