We use AI. Here is exactly how — and how we keep it safe.
TalkAlong is an AI-assisted practice tool built for people who may be vulnerable. That's a responsibility we take seriously — so we disclose every place AI is involved, keep a person in charge of every decision, and hold ourselves to Australia's recognised standards for safe and responsible AI.
The standards that guide us
We build and run TalkAlong guided by four reference points:
- AS ISO/IEC 42001:2023 — the Australian adoption of the international standard for managing AI responsibly: how an organisation should govern, plan, operate, evaluate and continually improve its use of AI.
- Australia's AI Ethics Principles — the eight national, human-centred principles for safe and responsible AI.
- Australia's Voluntary AI Safety Standard — the ten practical guardrails for organisations using AI, covering accountability, risk, data governance, human oversight, transparency and contestability.
- Speech Pathology Australia's 2024 position on AI — that AI tools should support, never replace, clinical judgement. TalkAlong is built as an adjunct, consistent with this.
Where AI shows up — and what it sees
Nothing in TalkAlong pretends to be a real person or a clinician. The helper is a clearly signposted AI avatar, and each AI service receives only what it needs:
| What it does | What it receives |
|---|---|
| The conversation — the helper's questions, encouragement and tailored practice | Text only — it never hears audio |
| Understanding speech — turning the participant's voice into text so practice flows | A live voice stream, processed in Australia in real time and never stored — used to understand, never to score |
| The helper's face — the on-screen practice partner | Only our own synthesised audio — no participant voice, name or photo |
| Goal pictures — optional images that illustrate practice goals | The goal's words and a broad age range only |
The full list of providers and where they process data is in our Privacy Policy.
What that means in practice
- A person always decides — never the AI. The app never delivers a clinical verdict or a right/wrong on someone's speech; the carer and speech pathologist do. That's meaningful human oversight, by design.
- You always know when it's AI. The helper is a clearly signposted AI avatar and the conversation is AI-written — nothing pretends to be a real person or a clinician.
- Safety is built into every AI turn. A fixed safety charter is attached to every single AI request: be warm and dignified, keep everything age-appropriate and safe, never give harmful or clinical advice, never invent facts, never ask for sensitive details, and respond gently if the person seems upset. A family can add extra limits; none can be taken away.
- We collect as little as possible. No participant audio or video is stored — text transcripts only. Speech is transcribed to understand meaning, not to grade it.
- Data is protected and stays private. Each family's information is kept isolated, encrypted in transit, and shared only with the trusted services needed to run the app — never sold, never used for advertising.
- We're upfront about the limits. Speech recognition is unreliable for atypical speech, so we deliberately never let it judge — protecting both fairness and the person's confidence. And TalkAlong is an adjunct, never a replacement for a speech pathologist.
Questions, or want the full statement? Email hello@talkalong.com.au. See also our Privacy Policy and Clinical disclaimer.
Be first to try TalkAlong
We're opening up early access to families, adults and clinicians. Join the list and we'll be in touch as places open.
- The daily practice that actually gets done — without being the therapist
- An Australian voice that adapts to the person's age
- For less than one therapy session a month; may be NDIS-claimable