Automated Decision-Making (Safeguards and Transparency) Bill 2026

The short answer

Automated Decision-Making (Safeguards and Transparency) Bill 2026

The Automated Decision-Making (Safeguards and Transparency) Bill 2026 was introduced in the House of Representatives on 7 September 2026 by Kate Chaney, the independent member for Curtin, as a private member's bill. It would set mandatory safeguards for significant automated administrative action by Commonwealth government bodies. Its central rule is that certain decisions cannot be automated at all: the bill sets out six outright prohibitions, including automation that would likely produce unlawful action, automation of actions that require an evaluative determination — the exercise of a discretion, an evaluative judgement or the formation of a state of mind — and automation where a plain English and technical explanation of the system's outputs cannot be produced. No action could be automated until a risk assessment is published on a new Commonwealth Automated Tools Register, and assessments must classify each automation as high risk or not. High risk systems would carry extra duties: a human decision maker able to substitute a different action where the automated result is not correct or preferable, internal merits review within four weeks, external review by the Administrative Review Tribunal and judicial review, publication of rules, algorithms, weightings and testing, and review by the Commonwealth Ombudsman at least once every three years. The bill implements recommendations 17.1 and 17.2 of the Royal Commission into the Robodebt Scheme. It was at introduction stage when this article was prepared and had not been passed, assented to or commenced.

This is a federal system guide. State constitutions, parliaments and local-government laws can allocate comparable functions differently.

The useful question is not only “what is the rule?” but also “who administers it, which document controls it, and when might it change?” That distinction prevents an accurate general explanation from becoming wrong advice in a particular election, chamber or policy setting.

Evidence review

What the bill does

The Automated Decision-Making (Safeguards and Transparency) Bill 2026 was introduced in the House of Representatives on 7 September 2026 by Kate Chaney, the independent member for Curtin. It is a private member's bill, not a government measure, and it would set mandatory safeguards for significant automated administrative action by Commonwealth government bodies.

The bill's approach is to regulate automated decisions on three fronts at once: what may not be automated at all, what must be disclosed and assessed before automation is used, and what additional protections attach to systems classified as high risk. It also defines when an action counts as automated — where a computer program takes the action, or does something substantially and directly related to taking it, using the same test the Privacy Act applies. The explanatory material says that wording is intended to stop agencies recording a human as the decision maker in order to avoid the rules.

Introducing a bill is the first stage of the parliamentary process. It must pass both houses and receive royal assent before it becomes law, and it commences only on the date or dates its commencement provisions set. At the time this article was prepared, the bill had not been passed, assented to or commenced.

Evidence review

The six outright prohibitions

Clause 18 of the bill sets out six situations in which automation would be prohibited outright. The first covers automation that would likely produce unlawful action. The second covers an action that requires an evaluative determination — defined as a discretion being exercised, an evaluative judgement being made, or a state of mind being formed, a definition drawn from the National Disability Insurance Scheme Act 2013.

The third prohibition is notable in framing: automation would be banned where a plain English and technical explanation of the system's outputs cannot be produced. On that basis explainability stops being an aspiration and becomes a precondition for deployment, because a system that cannot be explained cannot lawfully be used for the action at all.

The remaining prohibitions cover automation that would place undue influence on a discretion, automation with unjustified human rights impacts, and automation that would cause serious damage to public confidence in government administration. Together the six provisions describe the boundary of what the bill's sponsor regards as legitimately automatable.

Evidence review

Registration, risk assessment and classification

Under the bill, no action could be automated until a risk assessment for it is published on a new Commonwealth Automated Tools Register. Each assessment would have to classify the automation as high risk or not. High risk classification is mandatory where the action could reasonably be expected to significantly harm a person, the community or the environment.

Where an action is automated and a law requires notice to be given, the notice would have to say that the action was automated and name the responsible officer and the available review rights. Clause 26 would require a meaningful explanation of the rules and facts applied, and of what would have changed the outcome — a requirement that makes it possible for a person affected to understand the basis of a decision and to challenge it.

The register and the classification duty are the transparency backbone of the bill. They convert an agency's internal choice about automation into a public record with a risk judgement attached, which is what makes external scrutiny and the ombudsman's review possible.

Evidence review

Extra duties for high risk systems

For systems classified as high risk, the bill would impose protections that go beyond disclosure. A human decision maker must hold the power to substitute a different action where the automated result is not correct or preferable. That is a direct answer to the complaint that automation had removed the ability of officials to correct a wrong outcome.

Internal merits review would have to be completed within four weeks, or two weeks where regulations prescribe a shorter period. External review by the Administrative Review Tribunal and judicial review would both have to be available. The rules, algorithms, weightings and testing behind the system would have to be published, so that the basis of decisions is open to inspection rather than treated as a black box.

The Commonwealth Ombudsman would maintain the register and review every high risk system at least once every three years. These are continuing obligations, not one-off approvals: the review cycle is designed to catch systems whose risk profile changes after deployment.

Evidence review

Why the robodebt royal commission is central

The bill implements recommendations 17.1 and 17.2 of the Royal Commission into the Robodebt Scheme, which reported in July 2023. According to reporting on the bill's introduction, the government accepted both recommendations but has legislated neither.

Between 2015 and 2019 the Robodebt scheme raised debts against more than 400,000 Australians through automated income averaging. The royal commission's findings made automated administrative decision-making a live legal question, and the recommendations sought to ensure that safeguards are in place before another scheme of that kind can be built.

The bill's sponsor told the House that automation is spreading across government without rules — that there is no requirement a human can override an automated decision and no requirement to test tools before use — and described the gap as a failure to act on the royal commission's recommendations. She cited automated tools affecting older Australians and people with disability, saying an automated tool had had the final say on aged care support packages and that human decision makers had been unable to override a result they believed to be incorrect.

Evidence review

How it fits with existing obligations

The bill is designed to go further than the transparency obligations already in train or in place. From 10 December 2026, Australian Privacy Principle 1.7 requires privacy policies to disclose the use of automated decision-making where a program solely makes a decision, personal information is used, and the decision could reasonably be expected to significantly affect a person's rights or interests. That obligation binds agencies and private entities alike.

Commonwealth agencies also operate under the Digital Transformation Agency's policy for the responsible use of AI in government, which since December 2025 has required accountable officials, published transparency statements, AI use-case registers, impact assessments and mandated foundational AI training for much of the Australian Public Service.

As the reporting on the bill notes, both instruments are disclosure and governance measures. Neither requires that a human can override an automated result, and neither creates a review right for the person affected. The bill is intended to fill exactly that gap by adding override powers and review rights to the existing transparency layer, and its prohibitions go further still by removing some decisions from automation entirely.

Evidence review

What happens next

The bill had been introduced and a second reading debate had begun at the time of writing. As a private member's bill it depends on the Parliament for time and support, so its progress is less predictable than that of a government bill. It may proceed to further debate, be referred to a committee for inquiry and report, or remain on the notice paper.

Because the bill touches how every Commonwealth agency makes administrative decisions, its definitions — particularly what counts as an evaluative determination and what makes an automation high risk — would be the focus of any scrutiny. Those definitions determine how much decision-making is removed from automation and how much is merely made transparent.

Only if the bill passes both houses, receives royal assent and commences would its prohibitions, register, classification and review duties have legal effect. Readers should treat the Parliament of Australia bills register and the explanatory memorandum as the authoritative sources for the bill's text and progress.

Common questions

Before you rely on the answer