Vụ bê bối nợ tự động Robodebt của Úc

Centrelink pays benefits fortnightly, and recipients must report their income each fortnight. The Australian Taxation Office, by contrast, records income by financial year per employer. From 2016, the Department of Human Services scaled up an Online Compliance Intervention, later nicknamed Robodebt, that matched these two data sets. Where annual ATO income did not line up with what a person had reported, the system divided the annual figure evenly across the fortnights of employment and treated the difference as an overpayment. Unless the person supplied records showing otherwise, the debt stood. The core flaw is simple: income is often lumpy. People on benefits frequently work irregular or seasonal jobs, so an even average says little about what they earned in any given fortnight. The scheme also shifted the burden of proof onto recipients, many of whom were young, poor, unwell or no longer had payslips from years earlier. Debts were sometimes referred to private collectors, and a 10 per cent recovery fee was added in some cases. In November 2019, in the Amato case, the Federal Court made orders by consent recognising that a debt based on income averaging was not validly raised, and the government stopped relying on averaging alone. A class action, Prygodicz v Commonwealth, settled in 2020 and was approved in 2021; the settlement was valued at about A$1.8 billion, including refunds, wiped debts and roughly A$112 million in compensation. A Royal Commission, led by Catherine Holmes, reported in July 2023. It described the scheme as crude and cruel, found it was neither fair nor legal, and found that concerns about its lawfulness had not been acted on. It recommended reforms including a consistent legal framework for automated decision-making. A common misconception is that the problem was a software bug. The software largely did what it was designed to do; the design and the legal basis were wrong.