Australia Robodebt: Automated Debt Issuance Without Human Review
Australia's Robodebt automated the calculation and issuance of welfare debts without human review, unlawfully raising approximately A$1.76 billion in debts against more than 433,000 people; the roughly A$751 million actually recovered was later repaid under a Federal Court settlement.
What happened
From 2015 to 2019, the Australian government operated the Robodebt scheme, which used deterministic software to automatically calculate welfare debts from averaged income data and then issued debt notices against citizens without the human review that had previously checked each determination. The system that computed the debt possessed the authority to issue the debt notice directly, removing the authorization check from the path. This resulted in more than 433,000 people being wrongly pursued for welfare debts, with approximately A$1.76 billion in debts unlawfully raised against them; roughly A$751 million that had actually been recovered was repaid under the 2020-21 Federal Court settlement, alongside later compensation settlements. A 2023 Royal Commission found the scheme to be crude, cruel, and unlawful.
What the agent did
The automated Robodebt system issued debt notices against named citizens without requiring human authorization, combining the debt calculation and issuance functions within the same unreviewed automated process.
The irreversible effect
More than 433,000 citizens had unlawful debt notices raised against them and entered debt recovery processes; approximately A$1.76 billion in debts was unlawfully raised, and the roughly A$751 million actually recovered was later repaid under the 2020-21 Federal Court settlement. The money was largely restored, but the non-monetary harm was not: the Royal Commission documented severe distress among people pursued for debts they did not owe, including suicides that families attributed to the scheme.
Root cause
The human approval gate that previously authorized debt determinations was removed from the issuance pathway; the system that calculated the debt was not denied access to issue it, and there was no requirement for a named officer to review and authorize each debt before issuance, violating segregation of duties and approval-gate principles.
How a maker-checker control would have refused it
MakerChecker would issue refusal "skill_not_granted" to block the calculator system from holding the debt-issuance grant, and refusal "high_risk_requires_gate" to force debt issuance through a governed flow requiring explicit approval by a named review officer before the debt notice is issued; both refusals would prevent unreviewed issuance and ensure every debt carries an audit trail of who authorized it.
Runnable reproduction
This incident ships as a runnable scenario in the open-source repository. Point the enforcement engine at the policy and watch the action get refused, with the refusal written to a signed audit record.
examples/australia-robodebt-automated-debt-recovery
Accuracy and corrections
This entry describes a publicly reported incident and is compiled from the primary sources listed above. Where an account is a legal allegation rather than an established finding, the entry labels it as such. Summaries can still contain errors. If you can document a correction, email hello@makerchecker.ai and we will review and correct it, with the change noted, within 14 days.
See it for yourself
Reading is one thing. Watch it block an agent.
One command starts the demo: an agent stopped from signing off its own work, and the signed evidence file an inspector can check for themselves.
Designed against the rules your auditors already enforce.