The AI Assessment Framework Enters a New Phase

More teams are experimenting with AI. Some are deploying chatbots, others are using AI to support decision‑making or automate routine tasks. The NSW Government has recognised that this shift needs clearer guardrails and has updated its AI Assessment Framework to reflect how AI is actually being used today.

What is the AI Assessment Framework (AIAF)?

The AI Assessment Framework (AIAF) is the NSW Government’s mandated approach for assessing the risks and responsibilities that come with using AI systems. It’s designed to help agencies understand the impact of an AI use case and identify what safeguards need to be in place before that system is deployed.

Importantly, the AIAF applies to any use of AI, whether it’s a simple internal productivity tool or a public‑facing system that interacts with residents. It also applies across the entire lifecycle of an AI system, from early ideas and procurement through to deployment and ongoing operation.

“While designed for government, the questions the AIAF asks, about privacy, accountability, fairness and security, are the same ones boards and regulators are increasingly asking of private organisations” commented Gavin Hall, Business Technology Consultant, Diamond IT.

Rather than focusing only on how an AI system works, the framework prompts agencies to consider broader questions around privacy, fairness, accountability, accessibility and cyber security.

Who mandates it, and how is it used?

The AIAF is mandated by the NSW Government and overseen by the NSW Office for AI. All NSW Government agencies are required to use it for every AI use case.

Agencies complete the assessment themselves, but the results matter. The framework classifies each AI system by risk level: low, medium, high or critical, and automatically determines what additional reviews or safeguards are required.

Low‑risk use cases can move forward quickly. Higher‑risk systems, particularly those that affect service delivery or interact with the public, are flagged for deeper review and may be escalated to the NSW AI Review Committee for independent oversight.

This risk‑based approach mirrors how many enterprises are starting to evaluate AI use cases internally, moving away from one‑size‑fits‑all policies toward proportional governance.

What’s new in the latest AIAF update?

The biggest change in the current version of the AIAF is how practical and usable it has become.

The previous framework was detailed but time‑consuming, often requiring significant subject‑matter input and lengthy self‑assessments. The updated version introduces a structured, automated risk‑scoring approach that surfaces the right questions at the right time.

As Daniel Ward, Principal Advisor for Enterprise AI Risk and Innovation at the NSW Office for AI, puts it:

“The new framework increases usability without losing any depth that the original framework had. All that knowledge is still retained within the AIAF — it’s just surfaced at the appropriate time.”

Low‑risk assessments can now be completed in as little as 15 to 30 minutes, while higher‑risk systems are automatically identified and directed toward additional assurance processes. The framework has also been updated to explicitly account for newer capabilities, including generative and agentic AI.

What makes the updated AIAF better?

For government teams, the new framework removes a lot of friction from early AI exploration while still setting clear expectations around accountability and risk.

It allows agencies to:

• Assess AI use cases earlier and more consistently
• Apply safeguards proportionate to the actual risk involved
• Reduce ambiguity around when additional reviews are required
• Build confidence that AI systems are aligned with public trust expectations

The NSW Government has been clear about its intent. In announcing the update, Minister for Customer Service and Digital Government Jihad Dib said:

“AI can transform government services, but we have a responsibility to use it safely. This framework ensures agencies identify risks early and apply the right safeguards.”

Ethical considerations aren’t treated as a separate checkbox. Factors like bias, accessibility, human rights and transparency are evaluated as part of the core assessment, helping teams address issues while changes are still easy to make.

Why This Matters to All Industries

While the AIAF is mandatory only for NSW Government agencies, its impact extends well beyond the public sector.

Organisations that supply AI‑enabled products or services to government are increasingly expected to demonstrate how AI risks are identified, managed and governed across the full system lifecycle. Frameworks like the AIAF are already shaping procurement conversations, assurance requirements and vendor due‑diligence expectations.

More broadly, the framework reflects a shift many regulated and enterprise organisations are beginning to feel. As AI becomes embedded in decision‑making, automation and customer interaction, there is growing scrutiny around accountability, transparency, privacy, security and fairness, regardless of whether an organisation is formally required to follow a government framework.

For industry, the updated AIAF provides a clear signal of where Australian expectations are heading: faster AI adoption, paired with clearer, risk‑based oversight. Many organisations are already adopting similar approaches voluntarily to build trust, satisfy boards and regulators, and avoid retrofitting controls after systems are live.

Even for organisations outside government, the AIAF provides a useful benchmark for what “good” AI governance looks like in practice and where the bar is being set.

Diamond IT Can Support Responsible AI Adoption