When using automated decision-making systems government agencies must continue to comply with administrative law principles of legality, fairness, rationality and transparency, says the Commonwealth Ombudsman.
In updated guidelines released by the watchdog – an independent body that investigates complaints about the actions and decisions of Australian Government agencies – the ombudsman says the legality and fairness of decisions “must be preserved” when automating a decision-making process. “This means close and ongoing liaison with administrative law experts. This includes seeking expert external independent legal advice where necessary,” say the guidelines.
Privacy requirements and human rights obligations must also be adhered to.
The legal frameworks of administrative law, privacy and human rights will assist agencies in designing ADM systems “to ensure that key risks in automated decision making are avoided”, such as algorithmic bias, inaccurate decisions being produced and unclear reasons for decisions, say the guidelines – updated with the assistance of the Office of the Australian Information Commissioner and the Attorney-General’s Department.
The use of automated decision-making by Australian government agencies is long-standing practice, dating back to 2001. In 2023 in NSW alone, 275 public sector agencies reported using ADM systems – 198 within the state government sector, 77 within the council sector.
Guidelines have to be regularly reviewed and updated to reflect technological advancements.
The ombudsman’s 2025 guidelines include the following checklist summarising some of the key considerations for automation of administrative action:
- assess whether the system meets each of Australia’s AI Ethics Principles
- map whether the decision-making path involves the exercise of judgement or discretion and seek legal advice to determine whether automating those decisions would be lawful and appropriate
- identify the legislative authority for the action or decision and for automating the action or decision.
Agencies must also:
- undertake a privacy impact assessment
- seek assurance from any contractors that legislative requirements and best-practice principles have been followed
- undertake testing and verification of rules to ensure decisions are legal, accurate, fair and consistent.
When it is possible for an automated system to make decisions without the use of human judgement, the agency needs to nominate a position with ultimate responsibility for the decision, say the guidelines – such as a department secretary.
Addressing privacy, risks can be managed as part of the ADM’s design, says the ombudsman. Agencies have an obligation to take “reasonable steps to protect personal information from misuse, interference and loss, as well as unauthorised access, modification or disclosure”.
As well, agencies should have a response plan for potential data breaches that includes procedures and clear lines of authority, which can assist the department to contain the breach and manage a response.
ADM projects must establish appropriate governance frameworks and ensure that legal, policy and program areas are involved during the system’s development, say the guidelines. “Agencies should ensure that there is sufficient human oversight and that project decisions regarding automated systems are adequately documented.”
Transparency is critical for government accountability, says the ombudsman. In that regard, “people should be informed when automated systems are being used to make decisions that materially impact on their legal rights or other significant rights”.
Finally, the guidelines state that customers must be provided with an opportunity to dispute an administrative decision made by or with the assistance of an ADM system.
“If appropriately designed, automated systems can improve the quality and efficiency of government service delivery,” says the ombudsman, “and provide benefits such as improved consistency in decision-making and new service delivery options”.
Leave a Reply