Privacy & Responsible AI Policy

How Akrion One protects customer privacy and ensures the responsible use of AI on behalf of APRA-regulated tenant banks.

Version 1.0 · Effective April 2026 · Owner: Akrion One Platform Governance

ℹ️ This page summarises the full policy

The authoritative source is docs/AKRION_ONE_PRIVACY_RAI.md in the open repository. It is grounded in three frameworks: the Microsoft Responsible AI Standard, the 13 Australian Privacy Principles (Privacy Act 1988), and Australia's Voluntary AI Safety Standard, and aligns with APRA CPS 230, CPS 234, and IRAP.

1. Scope

The policy applies to every AI agent, data pipeline, and supporting Azure service operated by Akrion One on behalf of tenant banks (Acme, Pacific, Coastal Credit Union, and future tenants), across development, staging, and production environments. It does not replace each tenant bank's own privacy policy — every tenant remains an APP entity under the Privacy Act and is accountable for its own obligations.

2. Microsoft Responsible AI principles — at a glance

PrincipleHow Akrion One implements it
FairnessAnti-discrimination guardrails in system prompts; responses grounded in tenant policy & banking data — never demographic profiling. Quarterly bias review using the Foundry Evaluations SDK.
Reliability & SafetySix-layer AI safety stack — input sanitisation, threat detection, prompt guardrails, response verification, escalation logic, offline evaluation. Human-in-the-loop for loans > $200k, low confidence (< 0.7), and regulatory-sensitive topics.
Privacy & SecurityData sovereignty in Australia East / Australia Southeast, enforced via Azure Policy. Tenant isolation via Cosmos DB partition keys, PostgreSQL row-level security, AI Search filters, per-tenant ADLS containers. AES-256 at rest, TLS 1.3 in transit. No cross-tenant data use, no model training on customer data.
InclusivenessNatural-language interface — no banking jargon required. Adaptive communication style. Same AI capabilities for every tenant regardless of size.
TransparencyAI is disclosed at conversation start. Chain-of-thought logged. Confidence scores recorded. Standard financial disclaimers on every advice-shaped response.
AccountabilityShared Responsibility Matrix; full audit trail for every interaction; AI-specific incident playbooks in Sentinel; tenant notification within 24 h; quarterly Responsible AI review.

3. Customer privacy commitments

4. AI-specific safeguards

5. Shared responsibility

Akrion One operates the platform; tenant banks remain the data controller and accountable APP entity. The boundary is documented in the Shared Responsibility Matrix and the Azure Control Matrix.

6. Reporting a privacy or AI safety concern

Customers should contact their bank in the first instance. Tenant bank privacy and risk teams may escalate to Akrion at contact@akrion.one — see the contact section for response SLAs.

© 2026 Akrion AI Pty Ltd · akrion.one · Full policy on GitHub