Machine Learning-Integrated Medical Devices in Australia: Safety Defects and Regulation

Artificial intelligence (‘AI’) aspires to revolutionise health care, facilitating the provision of more rapid, accurate and personalised care. Yet AI models are only as good as the data used to train them and the parameters humans set to measure success. The risk of harm when poorly trained and inadequately tested machine learning-integrated medical devices are used on the market is extensive. This article scrutinises Australia’s product liability scheme, considering what algorithmic errors or failure modes might constitute a ‘safety defect’ and what defences manufacturers may seek to rely on. It challenges understandings of algorithmic opacity and grapples with the line between innovation and safety. Finally, it interrogates what is known about the regulation of such devices in Australia, raising questions about the application of the Australian medical device regulatory framework in this context.