Government faces criticism for opaque AI decision-making rules

The federal government is facing fresh criticism over its use of artificial intelligence and algorithms in public services.

Dr Catherine Williams, chief executive of the Centre for Public Integrity, warned that authorities are moving toward a repeat of the robodebt scandal.

Her organization released a report titled Computer Says No on Thursday. The document analyzes how agencies utilize automated decision-making tools.

Williams stated that Australians may never know if an algorithm influenced a decision affecting their lives. She described the current state of governance as lacking an appropriate framework.

The report outlines eight specific recommendations for reform. It calls for a mandatory audit of all government systems that use automation. It also proposes a public register to track these tools.

The authors suggest establishing an independent oversight body with the power to monitor and investigate government actions. This body would ensure compliance with new standards for algorithmic use.

The inquiry was triggered by concerns regarding the integrated assessment tool, or IAT. This system determines the level of home support available to older citizens.

Assessors have found that the IAT routinely underestimates individual needs and entitlements. The government has prohibited aged care assessors from overriding incorrect algorithmic outputs.

Williams said the team observed the evidence regarding the aged care tool with growing concern. She noted the absence of a robust structure to govern these systems.

She described the situation as a slow drift toward another major error in welfare administration. The organization hopes the report will inform any legislative framework the government proposes.

The attorney general's department has been drafting an automated decision-making framework. Public consultation on this framework ended in January 2025.

Attorney General Michelle Rowland did not provide a timeline for the completion of the draft. She stated that the government is committed to safe and responsible use of these tools. The aim is to deliver timely and efficient services to the community. The framework will implement recommendations from the royal commission into robodebt. A report from the commonwealth ombudsman published in October found that transparency advice from that commission was ignored during the rollout of the IAT. Rowland did not confirm if the new rules would apply retroactively. She said the framework will improve accuracy and consistency in administrative decisions. It will also embed safeguards including transparency obligations and human oversight. The government expects the system to lift productivity and provide faster services. Independent Member for Parliament Kate Chaney introduced a private member's bill in September. The bill seeks to create the first mandatory legal framework for automated decision-making. Chaney noted that the consultation process finished nearly two years ago. She said the attorney general's office has told her the work is ongoing. Chaney argued that maintaining pressure is necessary to prevent the issue from being dropped. She observed that government departments currently make decisions with little consistency or transparency. While automation can help achieve cost savings, public trust in government remains low. Chaney emphasized the need for clarity on when decisions are automated and why. She said citizens need clear pathways to review those decisions. This level of transparency is what the public expects from government agencies.