The Australian Centre for Public Integrity has called on the government to require checks of all automated decision-making systems and maintain a public register of them. The centre’s head, Catherine Williams, warned that without transparent rules, the country could repeat the Robodebt scandal. The report “The Computer Says No” examines the use of these systems in government; its eight recommendations include creating an independent oversight body with the power to monitor how they operate, conduct audits and investigate.
Australia warned of risk of a new algorithm scandal
An algorithm in an aged-care support system prompted the report
The centre prepared the report amid growing concern about the use of an algorithm in the Integrated Assessment Tool (IAT). It is used to determine how much in-home support older Australians are eligible for. Reviews found that the system regularly underestimates people’s needs and the size of the support packages they receive. The government has also barred care assessors from manually correcting the algorithm’s erroneous decisions.
Catherine Williams said examples of the algorithm at work in aged care had exposed the lack of reliable rules for automated decision-making systems in government. The centre decided to publish its recommendations now so they could influence the government’s policy framework.
Rules have yet to be published
Australia’s Attorney-General’s Department is developing rules for the use of automated decision-making. Public consultation ended in January 2025. Attorney-General Michelle Rowland did not answer questions about when a draft would be ready and published. She said the government intended to use these systems safely and responsibly to provide people across the country with timely and effective services.
Rowland said the rules should take into account the recommendations of the royal commission into Robodebt. However, an October report by the federal ombudsman found that the government had ignored the commission’s transparency recommendations when developing and deploying the IAT. Rowland also did not clarify whether the new rules would apply to decisions made before they came into force. She said the policy framework should improve the accuracy and consistency of administrative decisions, establish safeguards including transparency requirements, and ensure automation is used lawfully and responsibly, with appropriate human oversight. The rules would also, she said, help raise productivity and make services faster and more reliable.
The Centre for Public Integrity’s eight recommendations include:
MP calls for rules to be enshrined in law
In September, independent MP Kate Chaney introduced a bill that would establish the first legally mandated regulatory system for automated decision-making in government. Chaney said the department’s consultation on the issue had ended almost two years earlier. After meeting with the attorney-general, she was told, she said, that work was ongoing. The MP called for continued pressure to ensure the rules were not delayed because of the complexity of the task.
Chaney also noted that government agencies make decisions about automation without a consistent approach or sufficient transparency. The government is looking for ways to cut public service costs, she said, and automating decisions could help. But people need to be told when a decision is made automatically and when that approach is not appropriate, given an explanation for the decision, and clearly shown how to seek a review. That, Chaney said, is what people expect from the government.
Dataist’s take: the debate is not only about algorithmic accuracy, but also about whether decisions affecting people’s access to government support can be scrutinised. The policy framework has yet to be published, while the federal ombudsman has already criticised the IAT’s lack of transparency. A mandatory register and independent oversight could make the use of these systems more visible to the public, but the government has yet to give a publication date for the rules.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X