Fact Check: Australian AI Governance

The Policy for the responsible use of AI in government is effective 15 December 2025 and sets mandatory requirements for non-corporate Commonwealth entities. Phase 1 requirements apply from 15 June 2026; Privacy Act ADM transparency obligations (APP 1.7-1.9) commence 10 December 2026.

Partially overstated. The APS AI Plan says each agency appoints a senior executive as Chief AI Officer with milestones to July 2026, but this reads as a whole-of-APS expectation and implementation target rather than a statutory mandate in the same sense as the DTA policy.

Confirmed. Circular DCS-2026-02 Use of Artificial Intelligence by NSW Government Agencies (issued 30 July 2026) replaced DCS-2024-04 and mandates the AIAF together with the new NSW AI Operational Policy. A modernised AIAF was launched on 27 January 2026. The AIAF web page itself still cites DCS-2024-04.

Needs revision. NSW confirms the 16-question structure, but the updated explainer describes low, medium or high outcomes, while the AIAF page still refers to high or critical escalation, creating ambiguity.

Confirmed with nuance. In NSW, digital projects valued at $5 million or more go through the Digital Assurance Framework, and AI projects above $5 million receive additional central oversight to confirm AIAF compliance. In WA, projects over $5 million, with mid-range or higher residual risk, or funded through the Digital Capability Fund go to the WA AI Advisory Board, whose advice is non-binding. The mechanisms differ.

Inaccurate dating. The framework publication was first published on 5 March 2024 and last updated on 25 August 2026. No public source supports a January 2026 date for the framework or its $5 million criterion.

Confirmed. The QGEA Artificial intelligence governance policy (version 2.0.0, August 2026) is mandated, requires governance arrangements aligned with ISO 38507, and requires departments to prepare a Foundational AI Risk Assessment (FAIRA), or agreed equivalent, when evaluating AI solutions. The FAIRA framework document itself is non-mandated.

Overstated. Official Victorian guidance says the framework is still in development and refers to it "once released"; no public release had been found as at 28 September 2026. Current controls are the administrative guideline on generative AI (made 27 November 2024), the PROV AI Technologies and Recordkeeping Policy (29 February 2024), and related legal overlays.

Overstated. Guideline 13.1 (now titled Use of Generative AI and Large Language Model Tools, version 1.4, approved 17 February 2026) is an optional across-government guideline, not a compliance instrument. The Office for AI publishes SA Government AI policies, standards and guidelines, but current official pages do not support a "nation's first" claim. The SA Government AI strategy and a Responsible Use of AI Policy are yet to be released.

Overstated. Tasmania's guidance (revision 1.4, approved by the Secretaries Board on 13 September 2024) says the state has no formal assurance framework and may use the NSW framework as detailed public guidance, which is weaker than formal integration.

Partly confirmed. The NT Government now publishes an Artificial Intelligence Policy with mandatory requirements for agencies, staff, contractors and service providers, and an AI Assurance Framework (November 2025) under which AI systems with a tangible impact must be assessed. No public source was found for an NT AI Advisory Board or Community of Practice.

Overstated. ACT has an AI policy and assurance framework, and the Human Rights Act 2004 is the correct rights instrument, but the policy asks AI system owners to consider human rights impacts rather than guaranteeing adherence.

Revised from marketing claim. Original assertion was a market-positioning claim requiring substantiation. Revised to qualitative value proposition without unverified superlatives.