Australia AI Ethics Principles (Department of Industry)
Australia's voluntary AI Ethics Principles (2019) enumerate eight principles that AI systems operating in Australia are expected to observe. Ledgix evidences each principle with a signed artifact from the ledger, policy snapshots, or the checkpoint chain.
Status: Full — every principle resolves to an artifact Ledgix produces today. Principle 3 (fairness) is supplemented with demographic bias audits for tenants that opt in to subject_context tagging.
Scope
The principles apply voluntarily to any organisation designing, developing, or deploying AI in Australia. They were published by the Department of Industry, Science, Energy and Resources and form the basis of the National AI Ethics Framework. Coverage spans wellbeing, human-centred values, fairness, privacy, reliability, transparency, contestability, and accountability.
Controls covered
| Field | Type | Required | Description |
|---|---|---|---|
| AU-AI-P1 | policy_snapshots / events_jsonl | Human, societal and environmental wellbeing | Policy-scoped actions with denied-harmful outcomes demonstrate the wellbeing lens. |
| AU-AI-P2 | events_jsonl | Human-centred values | Per-decision value-based reasoning referencing policy text. |
| AU-AI-P3 | events_jsonl | Fairness | Aggregate decision distribution baseline plus bias audits for tenants opting in to demographic tagging. |
| AU-AI-P4 | events_jsonl / key_history | Privacy protection and security | Data-access decisions captured with governing policy and cryptographic custody records. |
| AU-AI-P5 | checkpoint_chain / events_jsonl | Reliability and safety | Continuous integrity record plus reliability baseline via confidence distribution. |
| AU-AI-P6 | events_jsonl | Transparency and explainability | Per-decision reason, citations, and evidence chunks. |
| AU-AI-P7 | events_jsonl | Contestability | Per-decision identifiers allow contestation workflows to locate the exact action. |
| AU-AI-P8 | events_jsonl | Accountability | AI and human actors identified per decision via agent_id and human_principal. |
Evidence types referenced
- policy_snapshots — rights-respecting policy text, signed and versioned.
- events_jsonl — per-decision record with reason, citations, and accountability metadata.
- checkpoint_chain — continuous integrity record supporting reliability claims.
- key_history — signing key lifecycle for cryptographic custody.
Known gaps (if any)
None — every principle resolves to an artifact Ledgix produces today. Demographic bias audits require tenants to tag events with subject_context for stratification.
Audit pack workflow
Export an evidence ZIP for this framework from the admin console's Evidence Exports panel by selecting Australia AI Ethics Principles (Department of Industry) and a time window. Each control's evidence_locators[] in the included framework_mapping.json points to the corresponding file in the ZIP.
References
- Framework mapping JSON:
vault/internal/compliance/frameworks/australia_ai_ethics.json - Canonical source: Australia's AI Ethics Principles — industry.gov.au