Security & compliance
Built to a standard you can hand to a practice manager and their insurer.
In-app tracking is necessary but never sufficient. Everything below is tracked as live status inside the console — with an owner, a state and a review date.
Legal baseline
The regulatory floor.
- Privacy Act 1988 (Cth) & the APPs
- Health information is sensitive information under the Act, held to a higher bar than ordinary personal information — APP 3, APP 6 and APP 11 in particular.
- Notifiable Data Breaches scheme
- Mandatory OAIC and individual notification for eligible breaches, with the 30-day clock run as a first-class workflow rather than a PDF in a drawer.
- My Health Records Act 2012
- Applies the moment the platform connects to My Health Record. Not in scope until it does.
- State & territory health records law
- Retention periods vary by state; the retention engine enforces them rather than relying on someone remembering.
Technical baseline
ACSC Essential Eight.
Tracked as eight discrete controls, each with a maturity state and an accountable owner.
- E8·01
Application control
- E8·02
Patch applications
- E8·03
Configure Office macro settings
- E8·04
User application hardening
- E8·05
Restrict administrative privileges
- E8·06
Patch operating systems
- E8·07
Multi-factor authentication
- E8·08
Regular backups
Architecture
Decisions that are painful to retrofit.
- Australian data residency
- Production data and backups held in Australian regions.
- Encryption everywhere
- TLS 1.2+ in transit, encryption at rest including backups.
- Row-level security by practice
- Isolation enforced at the database keyed to practice ID and role — never by UI filtering.
- Immutable audit logging
- Append-only from day one. Updates and deletes are blocked at the database, for admins too.
- Server-side model calls only
- Every AI call runs server-side so it can be logged, rate-limited and kept out of the browser.
- Sub-processor review
- Formal third-party risk review for each integration: PMS, claiming gateway, SMS provider, model provider.
AI trust & safety
What the model is never allowed to do.
A hard-coded boundary, enforced in product and reviewed as a compliance control — not a guideline.
- ✕Release patient information to any external party
- ✕Cancel, move or create a clinical appointment without confirmation
- ✕Submit or resubmit a Medicare claim
- ✕Send any patient-facing message that has not been read by a person
- ✕Rank, flag or prioritise patients by clinical risk or urgency
External trust artefacts
Proof that lives outside the product.
- Independent penetration test
- Budgeted before the first practice goes live with patient-adjacent data.
- Plain-English DPA
- A data processing agreement a practice manager can actually read.
- ISO 27001 / SOC 2 path
- A public roadmap in year one; certification as the book of practices grows.
- Insurance
- Professional indemnity and cyber liability that explicitly covers errors arising from automation.
