HTI is working with government and the private sector on processes to ensure AI systems are trustworthy and reliable.
AI Assurance
Ensuring AI is trustworthy and reliable
Artificial intelligence is increasingly central to how organisations operate and make decisions big and small. But how do we know the AI we rely on is safe, fair and produces accurate outputs? Will our privacy be protected? Is it trustworthy? Is it inclusive or does it reinforce existing biases? Does it comply with existing accountability, human rights and other laws?
For many years, we’ve had systems that aim to ensure that products and services meet basic safety, reliability and other legal standards. There is an increasing need to apply similar processes to the deployment of AI, so people can know they are trustworthy and reliable.
HTI is developing and reviewing AI assurance systems for government and the private sector, with a view to promoting safe AI.
HTI’s review of the NSW AI Assurance Framework
The NSW Department of Customer service engaged HTI to conduct a significant, independent expert review of the NSW AI Assurance Framework (the Framework) between June and October 2023.
NSW is the first jurisdiction in Australia to implement an AI-specific assurance framework, which came into force in March 2022. Since then, it has been mandatory for government agencies using AI systems to conduct a detailed self-assessment using the Framework, and to seek the approval of the independent NSW AI Review Committee if certain conditions are met. The Framework has set the tone for other states and territory governments, as well as the federal government, who agreed in June 2023 to pursue a nationally consistent approach to AI assurance.
HTI’s work seeks to improve the relevance, effectiveness and usability of the Framework to support the safe and responsible use of AI across the NSW Government. HTI's review included a deep analysis of current and evolving legal obligations, emerging international standards for AI assurance, extensive stakeholder feedback and a user-centric review to highlight how ease-of-use might be improved.
The review found that the Framework is seen as both essential and deeply valuable to a broad set of stakeholders, including past and future users. HTI’s analysis revealed opportunities to clarify the scope of Framework and make it more approachable, using updated definitions, examples and an easy-to-use ‘triage tool’. We identified opportunities to improve the form of the Framework to support greater collaboration across diverse teams and recommended new documentation requirements to bring the Framework `in line with international standards.
AI Assurance for the private sector
Tools like the NSW AI Assurance Framework are not just important for government. The corporate sector is also seeing a demand to adopt similar assurance frameworks or templates to certify that AI technologies are safe and reliable.
The vast majority of Australian organisations are using or exploring how they can use AI to add value to customers, drive productivity or increase efficiency, a trend that is accelerating since the public release of a series of highly capable generative AI tools.
Corporate leaders need to improve the governance of AI used by their organisations, in the form of suitable strategies, governance systems, new forms of expertise and human-centred cultures around AI.
HTI’s 2023 report, The State of AI Governance in Australia, provides a timely overview of how organisations are approaching the governance of AI in Australia today.
Corporate Frameworks
In November Atlassian released its Responsible Technology Review Template as a guide for companies using and developing AI and other tech to avoid harmful, unintended consequences.
HTI provided Atlassian with input and advice, particularly on the applicability of emerging international technical standards. The tool acts as a framework to mitigate AI-related unintentional bias and other harm, and to ensure AI use is accurate, safe and effective.
HTI was really pleased to work with Atlassian on this initiative.
This is a great example of how our partners are pushing the frontiers of human-centred AI in a really practical way – by designing an easy-to-use framework that any team in any industry can use when they are working with AI technologies. - HTI Co-Director Prof. Nicholas Davis