Since Circular DCS-2024-04, NSW agencies must apply a 16-question AI Assessment Framework, with high-risk systems going to an AI Review Committee. An August 2026 audit of 10 agencies found 13 of 15 projects assessed but wide governance gaps.
Details
Promoter
NSW Department of Customer Service (Digital NSW); AI Review Committee
Period
2022–2026
Keywords
ai assurance, risk assessment, government, audit
Description
The NSW AI Assessment Framework (AIAF) evolved from the 2022 NSW AI Assurance Framework. Under Circular DCS-2024-04 it is mandatory, alongside the AI Ethics Policy, for all NSW Government use of AI across the system lifecycle. Agencies complete 16 assessment questions, identify risks and mitigations, and must submit high- or critical-risk systems (or those with high residual risk) to the NSW AI Review Committee via the AI Secretariat. Projects above A$5 million, or funded from the Digital Restart Fund, receive additional central oversight through the Digital Assurance Framework. The Audit Office of NSW examined 10 agencies with significant AI use in its report 'Internal controls and governance 2026' (17 August 2026). All 10 kept an AI register, but only 4 recorded formally assessed AI systems. Of 15 AI projects reviewed, 13 had been assessed under the framework and 4 lacked a cyber risk assessment. Only half of the agencies had an AI strategy or agency-level AI policy, and just one considered AI in procurement documents. Transport for NSW's seatbelt-enforcement system, where three human reviewers approve every potential infringement, was cited as good practice. The Auditor recommends that agencies assess all required AI solutions against the framework by 30 June 2027. Lesson: a mandatory gate produces assessments, but registers, strategy and cost tracking lag without follow-through.
Read the full analysis: https://www.digital.nsw.gov.au/policy/artificial-intelligence/ai-governance-assurance-and-frameworks/nsw-ai-assessment-framework
Implementation
Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.
Do you run this practice?
Claim it —
verified implementers get a public contact pathway and can propose corrections.
Data sources
Where this practice's information was retrieved from, and when.
Australia's Digital Transformation Agency piloted a draft AI impact-assessment framework with 21 volunteer federal agencies, testing whether it caught risks …
The UK's Algorithmic Transparency Recording Standard requires central government bodies to publish structured records of algorithmic tools that significantly influence …