Coming into effect in March 2022, the NSW AI Assurance Framework assists agencies to design, build and use AI-enabled products and solutions. Consistent with the NSW Government AI Ethics Principles, the Framework is designed to help agencies identify risks that may be associated with their projects.
Who should use it
From March 2022, the AI Assurance Framework will be required for all projects which contain an AI component, except if:
- It uses an AI system that is a widely available commercial application, and
- The solution is not being customised in any way or being used other than intended.
When it should be used
The AI Assurance Framework is intended to be used at all major points of an AI project’s lifecycle – from start to finish. Project teams should continually assess their projects against the Framework to ensure their project abides by the ethical principles.
As best practice, project leaders and policy makers should review the Framework and consider each of its questions as part of the first steps in any AI project design process, and as part of any review of an existing AI solution.
How to use it
Project teams using AI should consider the questions under each of the Framework’s five categories, capturing their responses as they progress. The Framework also requires project teams to make decisions along the way, including decisions on how or if a project should progress.
Project teams will be required to seek approval from a responsible senior officer or submit the completed assessment to the AI Review Body.
This AI Assurance Framework assists project teams using AI to comprehensively analyse and document their projects’ AI specific risks. It also assists teams to implement risk mitigation strategies and establish clear governance and accountability measures.
It is part of a suite of materials for guiding NSW Government in uptake and use of AI solutions:
- AI Strategy: Outlines a vision for the use of AI in government, supported by clear commitments that will help build maturity and assurance in using this emerging technology.
- AI Ethics Principles: Sets out the definitions of AI and principles that all agencies deploying AI are required to adhere to.
- AI Assurance Framework: Provides clear requirements for agencies to design, build and use AI technology ethically and appropriately.
The AI Ethics Principles are mandatory through NSW Circular DCS-2020-04.
Under the AI Policy and AI Assurance Framework, agencies are required to abide by the following ethical principles:
- Community benefit - AI should deliver the best outcome for the citizen, and key insights into decision-making
- Fairness - Use of AI will include safeguards to manage data bias or data quality risks
- Privacy and security - AI will include the highest levels of assurance
- Transparency - review mechanisms will ensure citizens can question and challenge AI-based outcomes
- Accountability – decision-making remains the responsibility of organisations and individuals
Relationship to NSW Government policies
This Framework refers to other NSW Government policies, legislation, and tools that provide context and information relevant to AI, including the NSW AI Strategy and ethical principles. This Framework is intended to complement existing agency policies and standards.
It is important that subject matter experts are engaged to provide guidance on any organisation-specific policies and standards.
The Framework supports the existing NSW ICT Digital Assurance Framework and Digital Restart Fund.
Maintaining this Framework
This Framework is a living document. It will be regularly updated as AI evolves, opportunities and risks are better understood, standards develop, and AI maturity across NSW Government grows.
When to submit your project to the AI Review Body
From March 2022, project teams will be required to submit their completed AI Assurance Framework assessments for review if the project:
- Uses AI, and
- Is funded under the Digital Restart Fund, or exceeds an estimated total cost of $5million.
Other projects that will require review are projects that:
- Use an AI system, and
- Identifies residual risks (after mitigations) during assessment against the Framework which are mid-range or higher.
The AI Review Body is currently in development. More information about the process for seeking review will be made available prior to March 2022.