How Kookaburra stands against your 10 AI obligations
Where Kookaburra clears the AI-tool obligations for you — and where it arms you to meet the rest, with ready-to-use wording inside.
AHPRA · Privacy Act (APP 1.7 & APP 8) · TGA (January 2026 guidance) · Disclosure deadline 10 December 2026
Australian privacy law changed on 10 December 2024, and the grace period for disclosing automated decision-making ends 10 December 2026. AHPRA has set clear obligations for practitioners using AI, and in January 2026 the TGA clarified which AI tools are regulated as medical devices. Here is exactly where Kookaburra — a non-clinical, Australian-resident practice-intelligence tool — sits against each one.
What Kookaburra is (and isn't)
- Reads only scheduling, billing and practice-management data — appointments, invoices, and patient/practitioner records. Read-only.
- Never reads or stores clinical notes, diagnoses or treatment content (built-in AHPRA s.133 guard).
- The AI reasons over de-identified data (opaque IDs); patient names and contacts are revealed only on your explicit, audited request.
- Runs on Anthropic Claude via AWS Bedrock in Sydney — your practice data stays in Australia and is never used to train AI models.
- Never writes to your PMS or patient records, and suggests no diagnoses or treatments.
| # | Obligation | Kookaburra's position | Status |
|---|---|---|---|
| 1 | Clinical accountability documentedAHPRA | Produces no clinical output and writes nothing to a patient record. Its analysis explicitly avoids diagnoses, health outcomes and treatment efficacy (AHPRA s.133). | Not applicable |
| 2 | Patient informed consentAHPRA | Disclose Kookaburra as a practice-data processor in your intake. The AI sees de-identified data by default, which keeps this light — wording provided in this pack. | You + we support |
| 3 | AI scribe assessed vs TGA thresholdsTGA · Jan 2026 | Not a scribe and suggests no diagnoses or treatments, so it falls outside TGA medical-device regulation. | Not applicable |
| 4 | Patient data not in public AIAHPRA · Privacy Act | No public LLM is ever used. Runs on Anthropic Claude via AWS Bedrock (Sydney), data is not used for training, and the AI reasons over de-identified data — never clinical notes. | Cleared |
| 5 | Privacy policy updated for automated decision-makingPrivacy Act · APP 1.7 | Kookaburra informs business decisions, not decisions affecting patient rights or interests. Ready-to-paste disclosure text provided in this pack. | You + we support |
| 6 | Cross-border data disclosurePrivacy Act · APP 8 | Your practice data and its AI analysis are stored and processed entirely in Sydney (AWS ap-southeast-2) — no overseas disclosure. Disclose ancillary vendors (email, payments) as you would any provider. | Cleared |
| 7 | PI insurance covers AI useAHPRA | Confirm AI-assisted practice management is covered with your insurer. A written tool description for them is in this pack. | Your clinic |
| 8 | AI tool limitations understood & documentedAHPRA | A plain-English description of what Kookaburra does, its data scope and its limitations is provided in this pack. | We provide |
| 9 | Workplace AI governance policyAHPRA | List Kookaburra as an approved, Australian-resident, non-clinical tool in your staff AI policy. A starter template is in this pack. | You + we support |
| 10 | AI use transparently communicated to patientsAHPRA | Because Kookaburra is non-clinical (it informs business operations, not patient care), what you need to disclose is lighter. Suggested patient-facing wording is in this pack. | You + we support |
The wording, ready to paste
Drop these straight into your privacy policy, intake forms, patient materials and staff policy. Replace anything in [brackets] with your clinic's details.
1 · Privacy-policy disclosure (Automated decision-making · APP 1.7)
2 · Patient transparency statement
3 · Intake / consent wording
4 · Description for your professional-indemnity insurer
5 · Tool limitations — for your records (AHPRA #8)
Data scope: appointments, invoices, and patient/practitioner records (read-only). It does not access clinical notes, diagnoses or treatment content.
Known limitations:
- It is business analytics, not clinical advice, and not a substitute for professional, financial or legal advice.
- Benchmark comparisons are indicative, not standards of care.
- Outputs are only as accurate as the data in your practice-management system.
- AI-generated narrative should be sanity-checked before you rely on it for a decision.
6 · Staff AI policy — starter template
Approved tools. Kookaburra Health — an Australian-resident, read-only, non-clinical practice-analytics tool — is approved for practice-management analysis. [List any other approved tools.]
Prohibited. Staff must not enter identifiable patient information into public AI tools (e.g. ChatGPT, Gemini, Copilot) or any tool that has not been approved in writing.
Consent & transparency. Patients are informed of our use of practice analytics through our privacy policy and patient materials.
Clinical responsibility. AI tools do not make clinical decisions. Practitioners retain full responsibility for all clinical care.
Escalation. Questions or concerns about an AI tool go to [name / role].
Review. This policy is reviewed at least annually. Last reviewed: [date] · Signed: [name].