Skip to content

Compliance posture

This page is for the person completing a security questionnaire about Tavrik, or deciding whether it can go near patient data.

How evidence reaches you Every request, policy decision and administrative action is written as an event to a hash-linked, tamper-evident chain. A dispatcher delivers events at least once to the systems the hospital already runs: the integration engine as HL7 ARV, a FHIR store as AuditEvent, the SIEM, and Microsoft Purview. Failed deliveries retry with backoff and park in a dead-letter queue an operator can replay. How evidence reaches you Tavrik Pty Ltd, trading as Tavrik AI © 2026 Tavrik Pty Ltd, trading as Tavrik AI. All rights reserved. Every action recorded Requests, decisions, admin changes Categories, never patient details Hash-linked chain Each event linked to the last Verify any range on demand Delivered to you At least once, in order, resumable Kill switch per connector Integration engine HL7 ARV · Mirth, Cloverleaf FHIR store AuditEvent resources over HTTPS SIEM Splunk · Elastic · Datadog · Sentinel Microsoft Purview Labels in, evidence out If a target is down Retry with backoff, then park Operator replays from the console retry, then park Evidence lands where your compliance team already looks. Nothing here asks them to learn a new console.

Tavrik maintains a control mapping against each framework below. A mapping walks the framework’s controls one at a time and records what Tavrik does about each, and what evidence demonstrates it.

A mapping is the working document used to prepare for an audit. It is the route to a certification rather than a substitute for one, and nobody independent has signed it yet. Read it as the answer we would give and the evidence we would show, which is what you need when scoping a pilot.

Where each framework stands is in the assurance roadmap below.

Framework Why a hospital asks Status
HIPAA Required for any US healthcare customer, and the basis of the healthcare pack. Mapped against the Security Rule. A Business Associate Agreement is signed by Tavrik Pty Ltd once readiness is independently assessed.
SOC 2 The default enterprise procurement gate. Mapped. Security is the committed Trust Services Category. Availability, Confidentiality and Processing Integrity are mapped for Type II.
ISO 27001 Required by many international and government buyers. Mapped. Certification follows the SOC 2 work.
Australian Privacy Act Governs patient information held by Australian hospitals. Mapped, including the Notifiable Data Breaches scheme. Tavrik Pty Ltd is a service provider handling data on your behalf.
EU AI Act Phasing in through 2026 and 2027. Mapped by role. Tavrik is not a foundation-model provider. Where a vertical pack forms part of a high-risk system, your hospital is the provider of that system and Tavrik supplies a component.

These are in place today and can be demonstrated live.

  • Role-based access control on the admin API, with a closed three-role vocabulary and a permission check on every request that fails closed on an unknown role or action.
  • Credentials hashed with Argon2id, never stored or logged in the clear, and shown exactly once at creation.
  • Tenant isolation at the database layer, not only in application code.
  • Encryption in transit with TLS 1.2 or later, 1.3 by default, and mutual TLS where your systems support it. Provider keys are held as ciphertext and unwrapped in memory at the moment of a call.
  • A hash-chained, append-only audit log covering every request, policy decision, redaction and administrative action. Reads of the audit log are themselves audited, so nobody browses it silently. Verification runs on demand over a single event, over a range, or over the whole chain end to end. From the point a deployment moves its chain into the database, that whole-chain verification covers one ordered history; a pre-cutover file-backed record on a deployment that ran two writers against one file is verifiable per event and per branch, not as one ordered history, and is retained with its state recorded in the first event of the chain that replaced it.
  • Every change through a reviewed pull request with automated gates.

Tavrik Pty Ltd handles data on your behalf, so several controls can only be satisfied on your side.

  • The policy and the data. You decide what your people may send to AI tools and which providers are permitted. Tavrik enforces the decision.
  • Identity. Operators sign in through your identity provider. Joiner, mover and leaver processes stay yours, and a revocation in your directory is what removes access.
  • The host and the network on the on-premises and dedicated cloud shapes: patching, backups and the perimeter around the machine.
  • Retention. Tavrik does not persist prompt or response bodies unless you opt in. If you opt in, the retention period is yours to set.
  • Your audit aggregation. Tavrik writes categorical audit events into the systems you already run. Their retention, monitoring and alerting live there.

For a pilot or a security review: the control mappings themselves, the architecture and data-flow documentation on this site, a live demonstration of audit-chain verification against real events, the third-party dependency inventory with licences, and documentation of exactly what is installed and where on each deployment shape.

Four items are scheduled rather than complete. Each is dated.

Independent penetration test. Scheduled ahead of the first pilot go-live, against the deployed configuration a hospital would run. An annual third-party test is the standing commitment after that.

SOC 2 Type I. The auditor is engaged and the observation period begins after the first pilot.

ISO 27001 certification. Follows the SOC 2 work, on the same control set.

Automated evidence collection. Access reviews, change logs and scan results are assembled by hand today. Automating that collection is scheduled alongside the SOC 2 observation period, which is when it starts paying for itself.

Organisational controls that an auditor samples on a schedule, such as security training records, formal access reviews, vendor risk management and tabletop exercises, are designed and begin operating with that observation period.

Write to security@tavrik.ai. Questionnaire responses are acknowledged within one business day.