A proportionate, harmonised approach to regulation of AI is required to strike a balance between enabling innovation and protecting society.
Author: Donna Bartlett
In response to the launch by the Department of Industry, Science and Resources (DISR) in June 2023 of the consultation process around how Australia should regulate Artificial Intelligence, 510 submissions were received with 447 published on DISR's website.
Submissions originated from a broad base including leading AI technology providers, representative bodies from the creative sector, research and educational institutes, participants in the financial services, health care and retail sectors as well as government agencies. Over 25% of submissions originated from individuals wanting to ‘have my say’.
A review of a selection of submissions highlights both cross-sector commonality and diverse views in the response to how the regulation of AI in Australia should be addressed.
The chosen regulatory approach needs to be pro-innovation
A consistent theme across the submissions is that Australia should regulate in a “pro-innovation” manner.
From the health sector, consistent with the approach in the proposed EU AI Act where health contexts are classified as ‘high risk’ AI applications, submissions recognise that a crucial regulatory balance must be achieved by weighing up risks to patient outcomes with the economic benefits fuelled by promoting innovation.
A suggestion from the finance sector is to encourage the use of AI-focused regulatory sandboxes that allow testing of innovative systems in a relaxed regulatory environment on a small scale, with organisations working with regulators to receive guidance. This approach is also proposed by regulators in the EU and UK.
Harmonisation and a sector specific approach
Submissions from the technology sector indicate a strong preference for Australia to achieve parity with international regulatory approaches to allow for the ease of cross-border collaboration.
The finance, government and creative sectors similarly raise international harmonisation but express the importance of a co-ordinated domestic regulatory approach. The submissions highlight the importance of consistency amongst domestic regulators and recommendations to achieve this include through a cross-agency AI governance model with representatives from regulators and industry.
A comparison of the concepts of regulating economy wide (horizontal regulation) with focusing on sector specific regulation for AI (vertical regulation) arises in a number of submissions.
Primary AI providers (such as OpenAI, Google and Meta) express mutual support for a sector specific, vertical approach. The basis of this support is the underlying concept that AI systems can be very different based on sectoral context and accordingly require deep sectoral expertise for effective regulation to be developed and managed. A common theme from industry is that imposing mandatory or voluntary regulation will require consideration of sector-specific AI applications balancing the risk factors.
Existing Australian laws apply to AI but there is a need for clarification
There is a common theme around the need for additional guidance on how existing laws apply or will be applied to AI.
This is consistent across sectors such as finance and health with advocation for an initial assessment of existing domestic law and regulations prior to any steps to introduce new legislation.
There is also general support expressed for an assessment to identify any gaps in the existing laws which do not adequately address sector-specific risks and challenges associated with the use of AI.
Intellectual property laws garner special mention. Concerns regarding the position on copyright and intellectual property in general, both from the perspective of training data, user inputs and outputs are a theme across submissions from a variety of sources.
AI providers express a need to address the legal uncertainty under Australia’s current copyright framework as compared to other jurisdictions where existing or recently introduced exceptions provide a degree of clarity in this area. Conversely, industry bodies from the creative sector express deep reservations regarding the inclusion of exceptions to Copyright for AI computation. In our next blog we will explore the copyright question in further depth.
The submissions to the discussion paper are to be used to inform government on regulatory and policy responses to AI and With the accelerating uptake of AI, particularly Generative AI and its broad application, we can expect to see ongoing focus on regulation, we can expect to see some or all of these common themes under further debate in the not too distant future.
Thank you to William Parker, Senior Analyst and Seren Ozdemir, Graduate for their contributions to this blog.