The NSW AI Assessment Framework

⚡ Quick Summary

The New South Wales Government’s AI Assessment Framework provides a structured set of 10 questions for public sector teams to assess the risks, impacts, and ethical considerations of AI systems. Tailored for real-world procurement and deployment, it maps risks to five public values and includes use-case categories, risk levels, and mandatory actions. This is a working-level tool—not a policy whitepaper—and is designed to support practical implementation of the NSW AI Assurance Framework.

🧩 What’s Covered

1. Purpose & Scope

Supports ethical and accountable AI deployment by NSW Government departments
Applies to both developed and procured AI systems
Aligns with the NSW AI Ethics Policy, NSW AI Assurance Framework, and Australia’s Digital Government Strategy

2. Use Case Classification

Assigns use cases to low, medium, or high impact categories
High-impact use cases include predictive policing, child protection, and biometric surveillance
Each category maps to required steps: e.g., internal review, peer consultation, expert panel review

3. The 10 AI Assessment Questions

Each is supported by:

Risk rationale
Implementation checklist
Decision guides for mitigation

The 10 questions are grouped under 5 public values:

A. Fairness

Does the system treat people fairly?
Are outputs fair and unbiased across demographics?

B. Transparency

3. Can people understand how the AI makes decisions?

4. Are AI uses disclosed to affected individuals?

C. Accountability

5. Who is responsible for outcomes?

6. Is there an appeals process or way to contest decisions?

D. Privacy & Security

7. Is data collection proportionate and lawful?

8. Is the system secure and auditable?

E. Wellbeing

9. Could the system cause harm, distress, or reduce public trust?

10. Has consultation occurred with impacted communities?

Each question also includes:

Assessment tips for frontline staff
Alignment with privacy law, procurement standards, and oversight practices
Notes for when to escalate to an external review

4. Supporting Materials

A checklist worksheet (in spreadsheet format)
Decision tree diagram to classify AI use risk levels (see page 9)
Case study guidance for applying the framework to existing systems
Optional links to expert panels and NSW’s AI Register

💡 Why it matters?

This is one of the most usable government-issued AI assessment frameworks available. It’s not aspirational—it’s operational. For agencies that need to implement AI oversight within existing procurement and service delivery processes, this tool delivers just enough structure without being burdensome. It’s also highly transferable beyond NSW, especially for other public bodies seeking risk classification plus ethics-by-design.

❓ What’s Missing

While it flags “use of AI in the private sector,” it’s not designed for business users
No embedded metric-based scoring or risk weightings—assessments are qualitative
Doesn’t address post-deployment monitoring or adaptive systems that evolve over time
Focus is strictly impact-based; no alignment to international standards (e.g. ISO 42001, OECD, NIST RMF)
Checklist is not integrated into a digital tool (currently spreadsheet-based)

👥 Best For

Public sector teams deploying or procuring AI systems
Policy units and digital transformation teams seeking ethical AI implementation methods
Government auditors and oversight bodies developing AI assurance practices
Legal, ethics, and data officers working on transparency and accountability practices
NGOs and civil society monitoring the responsible use of AI in government