I move:

This bill be now read a second time.

It's now more than a decade since the robodebt scheme was put in place and three years since the robodebt royal commission released its final report, yet still there are no rules around the use of automation in government decision-making. There's no requirement to tell an individual that an algorithm made a decision about them, there's no requirement that a human be able to override an automated decision and there's no requirement for automated tools to be tested before use.

This marks a failure by the Albanese Labor government—a failure made worse by its recent decision to automate vitally important decisions affecting older Australians and those living with disability. And it appears that this government is willing to wait for the next robodebt to happen rather than listen to community, to the experts, to civil society and to the robodebt royal commission and act now. I choose not to wait. I choose to act now.

This bill introduces for the first time strong, legislated safeguards and transparency requirements for the use of automation in government decision-making. I urge the government to consider this bill and to act with the urgency required, matching its actions to its words of strong condemnation of the robodebt scheme.

Between 2015 and 2019, the robodebt scheme used an automated system to raise debts against more than 400,000 Australians, many of them on low incomes and already doing it tough. The system automatically calculated and issued debts, often without a human even checking the file. And the result? People who'd done nothing wrong were told they owed thousands of dollars. Some spent years fighting to clear their names. Others never got the chance; the stress and shame of being pursued by their own government contributed to their deaths.

The Prime Minister and countless members of the Labor Party have stood in this place and condemned the robodebt scheme, vowing never to repeat the trauma and harm that it caused. The government accepted two recommendations of the robodebt royal commission: to introduce a legal framework to guide the use of automated decision-making in government, and to establish a body to monitor the use of automated decision-making in government. But the government has not implemented these recommendations. There are still no guardrails around the use of automation across government, and, as the memory of robodebt slowly fades from this place, automation is causing real problems once again.

For the best part of a year, an automated tool has been having the final say on aged-care support packages for older Australians. Incredibly, human decision-makers have been unable to override a result when they believe it to be incorrect. There's been concern and fear across our community. People don't understand how this automated tool works. People with decades of experience are arguing that the automated tool is producing decisions that are clearly incorrect. Older Australians are finding their support slashed, despite their declining condition.

And it's not just in aged care. Automation is increasing across the board. In the NDIS, the government is introducing automated tools to determine eligibility for support and then the level of support. Automation is increasingly being used in social payments, in immigration decisions and in environmental approvals, and concern is growing. Just last year, 27 civil society signatories joined ACOSS to write to the Attorney-General about the need for these safeguards for automated decision-making.

Now, there are legitimate reasons to use automation. Done well, it can process routine rules-based decisions faster and more consistently than a human alone, freeing up public servants to focus on complex cases where judgement and empathy matter the most; and it can support our government to deliver services like the NDIS and welfare at scale. But it needs safeguards, transparency and oversight. The government has done some public consultation on this issue, including a public survey that had a whopping 20 responses. That consultation process finished nearly two years ago, and we've seen no legislation. This isn't good enough.

Earlier this year, I consulted my community on the issue, receiving more than 760 responses to a detailed survey, alongside input from legal academics, administrative law experts and disability and welfare advocates. More than 80 per cent of respondents were uncomfortable with government using automated systems to make decisions about them, and nearly 80 per cent wanted legislated mandatory rules. That consultation became a policy paper, Fairer and faster government decisions, which I released in March.

Today, I'm turning this policy into legislation. This bill enacts the recommendations of the robodebt royal commission. It requires any Commonwealth department or agency automating a decision to comply with a set of safeguards and transparency requirements. It prohibits automation outright in circumstances where it would be unacceptable, such as where it would produce an unlawful outcome, where the decision requires genuine human discretion, where no explanation of the outcome can be given or where it would have an unjustified adverse impact on human rights.

Beyond that, the bill takes a risk based approach. Every automated action must be risk assessed before use. Where an assessment finds it to be high risk—meaning it could reasonably be expected to have a significant adverse effect on a person, the community or the environment—additional safeguards apply. A human decision-maker must be able to override the automated outcome. The system must be tested for accuracy and differential impact before it goes live, and stakeholders affected must be consulted. There must be a working complaints process, a genuine right to internal review decided within four weeks, depending on the risk of harm, and a right to external merits review and judicial review.

The bill also builds in transparency throughout the process. Individuals affected by an automated decision must be told it was automated and given a meaningful explanation of why the decision was made. Details of every high-risk system, including how it works and how it's been tested, must be published on a new Commonwealth automated tools register maintained by the ombudsman, who is also given real oversight powers to review these systems, report publicly and be properly resourced to do the job.

Automation is essential for government to deliver services at scale. As technology improves, the opportunities for efficiency are growing, but we simply cannot continue automating incredibly important and risky decisions without these safeguards and transparency around the process. I commend this bill to the House.

Governments increasingly use automated decision-making to deliver services more efficiently and consistently, but the robodebt royal commission found that unlawful use of automated decision-making harmed hundreds of thousands of Australians. Its final report recommended a consistent legal framework in which automation and government services can operate. That was in July 2023, and we're still waiting for that framework.

But, since November 2025, we have used the Integrated Assessment Tool on older Australians who need support at home. The tool is being used by clinicians, but the information that they collect is being fed into an algorithm which determines the final funding classification, and assessors can't override that outcome, even when their clinical judgement differs. After widespread concerns and complaints, the Commonwealth Ombudsman is now investigating the tool. It wasn't live tested before it was introduced, and now the government has commissioned a rapid review, having recognised that one size does not fit everybody.

In the NDIS, the I-CAN assessment tool has now been selected to inform calculation of participants' budgets. The government, again, has suggested that a single tool can foster consistency and fairness, but the reality exists and remains that a blunt and opaque instrument used against the diversity of disability and of need has been almost universally condemned by expert policymakers and by disability advocates.

Where automated decisions from government shape the daily lives of people with a disability or older Australians, we have to understand how those decisions are made. We have to have an avenue to challenge them, and we need capacity for fair and just review by a human. That's what this bill seeks to achieve. It contains a legislative framework for transparency, risk assessment, human accountability, review rights and independent oversight. It will assure that automation will serve us—not the reverse—and that humans have fair and transparent oversight of systems that should serve humans. I am pleased to second this bill, and I thank the member for Curtin for her leadership on this very important issue.

The time allotted for this debate has expired. The debate is adjourned, and the resumption of the debate will be made an order of the day for the next sitting.