Integrity thinktank urges audits of Australian government algorithms to avert another Robodebt

Australia’s Centre for Public Integrity is calling for mandatory audits and a public register of government automated decision-making, warning that weak oversight risks another Robodebt scandal. Its new “Computer Says No” report follows mounting concerns about an aged-care algorithm, while the government’s promised framework remains in development nearly two years after consultation closed.

The aged-care catalyst

Centre chief executive Dr Catherine Williams said concern about the government’s Integrated Assessment Tool helped drive the report. The tool uses an algorithm to determine older Australians’ home-support entitlements. The centre worries that people may not even know when automated decision-making—known as ADM—has been used in decisions affecting them.

The aged-care classification algorithm began operating on November 1, 2025, assigning funding levels for support at home rather than residential care. Hundreds of people have appealed its decisions, more than five times the previous year’s number, according to ABC reporting on the Commonwealth Ombudsman’s investigation.

In findings released on October 26, Ombudsman Iain Anderson said the department introduced the algorithm without meaningful risk assessment, evaluation or analysis. His investigation also found that transparency recommendations from the Robodebt royal commission had been ignored in developing and rolling out the assessment tool.

The Department of Health, Disability and Ageing disputed the claim that it had not tested or validated the algorithm. ABC reported that testing covered an original version created by a consultancy; the department subsequently modified it, and the version in use had no external validation.

It feels as though we are sleepwalking into another robodebt scenario, and that’s why we’ve decided to put this work forward now, in the hope also that it can help inform any framework that the government does propose in relation to ADM.

Dr Catherine Williams, Centre for Public Integrity chief executive, speaking to the Guardian

From one algorithm to government-wide scrutiny

The centre’s report makes eight recommendations. Its proposals extend beyond fixing the aged-care system, seeking a consistent way to identify and scrutinise automated decisions across government:

• A public register covering all government use of automated decision-making.
• An independent oversight body empowered to monitor, audit and investigate government use of these tools.

The aged-care dispute also exposes a question about who can correct an algorithm. Anderson recommended changing legislation to let senior assessors override incorrect decisions. The department noted that recommendation rather than accepting it. It accepted six of his 10 recommendations, including a comprehensive review of the assessment process and algorithm’s legal framework.

Before the Ombudsman’s findings, the government passed legislation allowing an escalation pathway for appeals in exceptional circumstances, instead of giving assessors blanket override permission. It also commissioned a review of the assessment system and algorithm, with a report due in the first half of 2027.

A framework without a completion date

The Attorney-General’s Department has been developing a broader ADM framework, with public consultation ending in January 2025. Attorney general Michelle Rowland did not give the Guardian a date for publishing a draft, or answer whether the framework would apply retrospectively.

Rowland said it would implement Robodebt royal commission recommendations and include transparency obligations and appropriate human oversight. She also said it would improve accuracy and consistency, lift productivity and deliver faster, more reliable services. These remain the government’s stated aims for a framework still being developed.

Independent MP Kate Chaney introduced a private member’s bill in September seeking a mandatory, legislated framework. She said departments currently approach automation with little consistency or transparency. Her demand goes beyond disclosure: people should understand why a decision was made and have clear routes to challenge it.