The track record of AI in government benefits

The history of AI in government benefit administration is not encouraging. Australia's Robodebt scheme — which used automated income averaging to generate welfare debt notices — was found by the Royal Commission to have been unlawful, caused significant harm to hundreds of thousands of people, and contributed to deaths. The UK's Universal Credit system has been repeatedly criticised for automated decisions that failed to account for individual circumstances. The Netherlands' child benefits scandal (the "toeslagenaffaire") involved algorithmic fraud detection that wrongly flagged thousands of families, many from ethnic minorities, and resulted in government collapse. In the US, states have used AI to cut Medicaid and disability benefits, with courts finding systematic errors in algorithmic assessments.

These are not isolated incidents — they reflect a pattern of automated government decision-making that prioritises efficiency over accuracy and that systematically disadvantages the most vulnerable people in the system.

Your rights when government AI affects your benefits

The right to human review: in Australia, decisions affecting your welfare payments can be reviewed by a human decision-maker within the agency (internal review) and then by the Administrative Review Tribunal (formerly AAT). These rights exist regardless of whether the initial decision was made by AI. Do not accept that an automated decision is final — it never is. The right to reasons: you are entitled to know the basis for any government decision affecting your benefits. Request reasons in writing. The reasons must be adequate to understand the decision and to appeal it effectively — they cannot consist solely of references to an algorithm or a score. The right to accurate information: if a government decision was based on data about you that is incorrect, you have the right to correct that data and to have the decision reconsidered. In Australia, the Privacy Act gives you the right to access and correct personal information held by government agencies.