--- title: "Trust Center β€” Hanah (AU)" description: "Hanah Trust Center (Australia): security and privacy controls, subprocessors, and live model disclosure for the Australian production deployment." source: https://hanah.health/au/trust/ --- # Trust Center Transparency for clinicians, practice owners, and institutional buyers. This page summarises how Hanah handles health information, which subprocessors are in scope for the production deployment, and how each security control is implemented. Status markers are a temporary editorial aid and will be removed once every item is production-verified. βœ… Done πŸ”§ Implemented bespoke Overview Controls Subprocessors Resources **Who this is for:** Clinics, health networks, privacy officers, and IT security teams doing due diligence on Hanah. **At a glance:** this Trust Center summarises Hanah's security controls and the subprocessors in scope for the production deployment. Speech-to-text and LLM inference run under enterprise terms that prohibit training on your content. Deeper evidence sits in our Security & Operations Handbook; the **Controls** tab describes how each control is implemented. ### Security controls Encryption, access control, network segmentation, logging, backup/DR, and AI-specific safeguards β€” with an honest status per line. View controls β†’ ### Subprocessors Who touches data, for what purpose, in which region, and what attestations we rely on. View subprocessors β†’ ### Resources & next steps Service status, how to request full policy packs under NDA, and model disclosure (API). View resources β†’[ ### Validation methodology How we measure transcription accuracy and documentation hallucination β€” one harness, one corpus, two evaluation layers (WER + LLM jury). The rules every published number is graded by. Read the methodology β†’](https://hanah.health/au/trust/wer-methodology/index.html)[ ### Validation transparency Per-recording transcripts, generated documents, and per-LLM juror verdicts against an authoritative ground-truth corpus. The full data, not just headline numbers. View report β†’](https://hanah.health/au/trust/transparency/index.html)[ ### Past results Every benchmark we have published, with the change against the run before it. Numbers are left exactly as they were measured on the day β€” including where the grading panel itself has since changed. Compare runs β†’](https://hanah.health/au/trust/benchmarks/index.html) Each row is a control in the usual ISO sense: **policy** (governance and documented procedure), **technical** (enforced by systems and configuration), or **vendor** (contractual and subprocessor assurance). Tags show the dominant evidence type; many rows combine more than one. Detail traces to our internal **Security & Operations Handbook**. Technical System and platform controls β€” infrastructure, application, cryptography, automation. Policy Organisation-level governance β€” accountability, signed policies, risk register, procedural evidence. Vendor Third-party reliance β€” DPA terms, subprocessor list, supplier certifications. ## Infrastructure & platform Infrastructure security Defines where Hanah runs for the regional deployment, how production is segmented from operations, and how encryption protects customer data between clinicians, our stack, and AI subprocessors. These are mostly technical controls with administrative backing in the Security & Operations Handbook. - βœ… Technical Policy **Region-resident production workloads** Primary application servers, databases, object storage, and backups for the deployment are located in-region. Administrative records describe regions and change approval; technical deployment pins resources to the regional project. - βœ… Technical **Encryption in transit** Public ingress uses modern TLS. Traffic to speech-to-text and LLM inference subprocessors uses TLS to the configured regional endpoints only. This is a technical control verified in architecture reviews and deployment config. - βœ… Technical Policy **Encryption at rest** Databases, object storage, and backup objects use industry-standard algorithms (e.g. AES-256-GCM) with keys managed per our cryptography policy. WAL-G backups add client-side encryption before off-site storage. - βœ… Technical Vendor **Edge DDoS protection and WAF** A managed edge layer fronts public hostnames, absorbing volumetric attacks and applying WAF rules for common web abuse. Provider-level protections and tight firewall posture remain in place behind the edge. - βœ… Technical Policy **Intrusion detection** Layered detection: the hosting platform provides network-layer IDS, while each VM runs fail2ban (SSH brute-force prevention), UFW with default-deny posture, and unattended security upgrades. Authentication and access events from the auth-proxy are shipped to a central Loki/Grafana stack so anomalous login patterns surface in dashboards and alerts. Policies assign triage ownership when alerts fire. Kernel-level runtime IDS for containers is a planned enhancement on top of this baseline. ## Product & application Product security How users authenticate, how authorisation is enforced in the product and API layer, and how we build and ship software. Mixes technical enforcement (JWT, RBAC) with administrative SDLC and review requirements. - βœ… Technical Vendor **Authentication and sessions** Clinicians authenticate through Firebase (OIDC/OAuth-style flows); every API path validates identity. Session lifetime and lockout behaviour align with our access-control policy. - βœ… Technical Policy **Role-based access control** Clinician, patient, and admin capabilities are separated in the data model and GraphQL layer so least-privilege is the default. Role definitions and review cadence are documented administratively. - βœ… Technical Policy **Mandatory MFA for privileged paths** Mandatory MFA is enforced for accounts that can change security-sensitive settings or access high-risk data, via Firebase Authentication. - βœ… Technical Policy **Secure development lifecycle** Changes flow through source control, review, and pinned dependencies. Security impact is considered for substantive features as an administrative gate; technical tools (Dependabot, etc.) support ongoing hygiene. - βœ… Technical Policy **Software bill of materials (SBOM)** Each release carries a machine-readable inventory of dependencies for vulnerability response. SBOM export is wired into CI as part of our supply-chain assurance programme. ## Data & privacy Data & privacy Retention, deletion, subject rights, and visibility into security-relevant events. These controls pair automated enforcement in the product and database with administrative procedures (DSAR handling, breach notification, quarterly reporting). - βœ… Technical Policy **Retention and deletion for ambient artefacts** Audio and transcript material defaults to a short retention window with automated deletion; clinical notes follow each organisation’s retention policy. Once integrated with your record system, the authoritative clinical record remains yours. - βœ… Technical Policy **Tamper-evident deletion log and quarterly reports** Retention-driven deletions are recorded in `deletion_log`, an append-only Postgres table with no `UPDATE` or `DELETE` grants to any role. Each row links to the previous one by SHA-256 hash chain (`row_hash = sha256(prev_hash β€– canonical(metadata))`), so altering or removing any earlier row breaks every subsequent hash and is detected by the in-database verifier `hanah_verify_deletion_chain()`. Companion audit tables (`ai_audit`, `secret_access_audit`) are also append-only. Quarterly customer-facing rollups are produced by `hanah_deletion_report_quarter(year, quarter)`. - βœ… Policy Technical **Data subject access and correction** Individuals can request access to or correction of their personal information; Hanah documents timelines, formats (JSON + human-readable exports), and escalation paths. Tooling supports export where data lives in Hanah systems. - πŸ”§ Technical Policy **Unified audit visibility** Security-relevant events (authentication, privilege use, configuration changes) are queryable for institutional customers and forwardable to your SIEM (for example Microsoft Sentinel) where contractually agreed. Configured per customer. ## AI & clinical safety AI security & compliance Controls specific to ambient AI: no secondary use of health data for training, human oversight before clinical reliance, measurement of quality (WER), and abuse-resistant prompting. Combines contractual third-party controls with product workflow and clinical governance. - βœ… Vendor Policy Technical **No secondary use for model training** Our speech-to-text and LLM inference providers' enterprise agreements prohibit training on customer content; Hanah does not operate an internal training pipeline on your health data. Administrative vendor reviews and technical configuration (region-locked endpoints) support the claim. - βœ… Policy Technical **Human-in-the-loop before clinical reliance** Generated notes require explicit clinician review and finalisation before they are treated as part of the clinical record. This is both a product control (workflow gates) and a policy commitment in the PIA. - βœ… Policy Technical **WER methodology and published cohort results** One harness, two evaluation layers, run on every grading pass: **Word Error Rate** against verbatim ground-truth transcripts (Layer A), and an **LLM jury panel** measuring both _coverage_ (does the note capture each pre-registered expectation?) and _hallucination_ (does the note contain claims the transcript does not support β€” i.e. content the model invented?). The [full methodology](https://hanah.health/au/trust/wer-methodology/index.html) is published, and [live results](https://hanah.health/au/trust/transparency/index.html) render directly from the harness's aggregate output. The corpus continues to graduate v1 (synthetic) β†’ v2 (role-play) β†’ v3 (real consent-based) β†’ v4 (joint with health-system partners). - βœ… Technical Policy **Prompt injection and unsafe-output monitoring** System prompts are server-controlled and inputs are bounded by the clinical workflow (no free-form chat surface). Provider safety classifiers run on every generation; when they trigger, the stream surfaces a content-filter finish reason that is captured in the per-call `ai_audit` row alongside a canonical `ai.content_policy` error code, latency, tokens, and a Tempo trace ID. Anomalous patterns surface in dashboards over that audit table. Cross-call behavioural anomaly detection for abuse patterns is a planned enhancement on top of this baseline. - βœ… Technical Policy **Multi-party consent capture in product** When the clinician clicks record, the product opens a consent dialog requiring explicit confirmation that consent has been obtained from the patient and any other participants in the session. Recording cannot begin until that check is acknowledged, and the confirmation is bound to the encounter alongside clinic policy referenced in the privacy impact assessment. ## Organisation & third parties Organisational security How Hanah the company governs security: accountability, policies, independent assurance, and integration with customer security tooling. These are primarily administrative controls; some have technical deliverables (e.g. SIEM connectors). - βœ… Policy **Security governance and risk register** A named executive is accountable for information security; risks are logged with owners and target dates and reviewed on a fixed cadence. This matches what buyers typically expect from an ISO-style management system, even at our current scale. - βœ… Policy **Board-approved policy pack** Formal policies cover access, cryptography, incident response, vendors, and change management. The pack is board-approved and held under version control; full content is available under NDA. - πŸ”§ Policy Technical **Customer SIEM forwarding (e.g. Sentinel)** Security events from Hanah-controlled systems are forwarded into your Sentinel workspace (or equivalent) for joint monitoring where contractually agreed. Configured per institutional customer. Subprocessors listed here are those that **may process or store Hanah customer data** in our production architecture. Regions reflect how we configure services for health workloads. Clinician sign-in may use global identity infrastructure where **no patient health information** is stored β€” see the Firebase row. Subprocessor Purpose Data categories Region Notes ### Vertex AI / Gemini β€” models in use The rows below are pulled live from Hanah’s public `/api/trust` registry (same contract as audits). Only entries with **vendor Google** and **deployment Vertex** are shown here β€” i.e. LLM inference routed through Google Cloud Vertex AI in the region configured for this deployment, not the consumer Gemini API or Google STT. Loading model list… If this fails to load (network or ad blocker), open the raw JSON: [](https://au.hanah.health/api/trust)β€” region is set via `data-trust-region` on the `` element. Operational transparency: a public **status** page (uptime and incidents) sits alongside the production `/healthz` readiness endpoint used by our own monitoring. [Service status](https://hanah.health/au/trust/status/index.html "Per-component health, 90-day uptime, and incident history for the AU deployment.") [Full model registry (JSON)](https://au.hanah.health/api/trust "Full public registry (all vendors / deployments). Vertex Gemini slice is also inlined under Subprocessors.") [Request full security pack under NDA](mailto:security@hanah.health?subject=Trust%20%2F%20security%20pack%20request) [Back to hanah.health](https://hanah.health/index.html) PhysiPal Pty Ltd (trading as Hanah) β€” . This Trust Center is a high-level summary. Contractual commitments, data processing terms, and subprocessors in force at signing supersede this page. Status emoji markers (βœ… πŸ”§) are temporary and will be removed after independent verification of each row.