The accelerating development of Artificial Intelligence (AI) brings with it the possibility of great change in human society, impacting how we live and work, improving productivity, supporting learning and offering new entertainment possibilities. ATSE’s report for the National Science and Technology Council on generative AI illuminates the risks and opportunities for Australia, as well as challenges such as balancing progress with human rights, data security and environmental impacts.
The potential impact of this rapidly developing branch of technological advancement necessitates thoughtful and responsive oversight and regulation, and ATSE commends the Australian Government for commencing with the Safe and Responsible AI Practices Discussion Paper. ATSE agrees that the level of regulation applied to AI systems should be proportionate to the risk that misuse or malfunction of the technology may have to individuals, industry and society. Many companies, both in Australia and internationally, are moving ahead of the Government by developing their own ethical AI policies, principles, standards and strategies. Independent oversight and guidance are important to ensure that AI systems are used appropriately and that bad actors can be identified and prevented. The regulatory environment must remain flexible enough to adapt to rapid changes in technology, supporting the sector to grow and develop, while minimising the risk of potential negative impacts.
ATSE commends the tiered model proposed in the discussion paper as fundamentally the right approach to regulating AI in Australia. ATSE’s recommendations in this submission focus on changes designed to support regulation that meets its intended aims, and identify factors to be considered if this framework is translated into law. The nature of data collected and analysed by AI models should also be more fully incorporated into the risk matrix to ensure appropriate use and storage of private or sensitive data. Regulation should address AI systems under development and those that store data offshore, and protect vulnerable people.
To this end, ATSE makes the following recommendations:
Recommendation 1: Ensure Australia’s AI regulatory environment is consistent with the OECD AI principles and can adapt to future AI technology development.
Recommendation 2: Include AI systems under development within the graded regulatory requirements based on the potential impacts of the AI system.
Recommendation 3: Assess the risk categorisation of AI systems under the proposed regulatory framework against real-world outcomes of the system, as evaluated by ethical AI experts and representatives of those impacted by the AI system.
Recommendation 4: Require reporting of aggregate data for medium risk AI systems to ensure these systems do not perpetuate or exacerbate systemic disadvantage.
Recommendation 5: Amend the definition of the proposed risk categories to incorporate the risk posed by data inputs to AI systems, in addition to the current categorisation based on outputs and impact.
Recommendation 6: Require AI systems operating in Australia to meet Australian regulatory requirements for privacy and data security regardless of the location of their servers.