The Benefits Scandal That Brought Down a Government
The Dutch tax authority used a risk model, with citizenship among its inputs, to choose which childcare-benefit claims to check. Tens of thousands of parents were wrongly treated as fraudsters, and the cabinet resigned.
When a model chooses whom to investigate, how do its choices turn into accusations?
What happened
To decide which childcare-benefit claims to check for fraud, the Dutch tax authority used a risk-classification model. According to Amnesty International, one of its inputs was whether the applicant held Dutch citizenship, so non-Dutch nationals received higher risk scores. Tens of thousands of parents were wrongly treated as fraudsters and told to repay benefits, with severe consequences for their finances and their families. By February 2025 the Dutch government had recognised more than 40,300 parents as affected.
Who spoke up
Parents. Those who had been wrongly accused pushed for recognition and compensation. On the day the government resigned, one mother told reporters outside parliament that the government had to take responsibility.
What changed
In December 2020 a parliamentary inquiry published its report, titled Ongekend onrecht (“Unprecedented injustice”), finding that fundamental principles of the rule of law had been violated. On 15 January 2021 the cabinet of Prime Minister Mark Rutte resigned.
In December 2021 the Dutch Data Protection Authority fined the tax administration €2.75 million for processing applicants’ dual nationality in ways it found unlawful, discriminatory, and improper. That data should have been deleted in 2014; in May 2018, 1.4 million people were still registered. A compensation and recovery scheme for affected parents followed.
What to keep straight
- The number of families affected depends on when it was counted: more than 20,000 in early 2021 reporting, and more than 40,300 officially recognised by 2025.
- The €2.75 million fine concerned the processing of dual nationality. The citizenship input in the risk model is documented by Amnesty International and the authority’s own investigation.
- The government argued that taking citizenship into account is not the same as treating nationality as a risk factor.
Talk about it
- One input in a model can shape thousands of lives. Who should review the inputs before a model is used?
- Being picked for a check is not the same as being found guilty. Where did that line get lost?
- What would a fair way of choosing whom to check look like?
Go further
We Are Living in the Effects of a Cause We Never Authorized
Borrowing the causal invariance principle from physics, this essay argues that AI is producing a structural violation of causal order in society, displacing workers and restructuring cognition before any deliberate social cause has authorized these effects.
Read the article Teach it Classroom Kits · Ages 12 and upWho Decides
A role-based circle debate on one real scenario: a school considering a tool that flags 'distracted' students. Every seat at the table gets a voice, including the ones usually left out.
View the kitSources
- Xenophobic Machines: Discrimination through unregulated use of algorithms in the Dutch childcare benefits scandal, Amnesty International (25 October 2021).
- Eindverslag onderzoek kinderopvangtoeslag overhandigd (final report of the parliamentary inquiry, in Dutch), Tweede Kamer der Staten-Generaal (17 December 2020).
- Boete voor Belastingdienst van 2,7 miljoen voor discriminatie toeslagenouders (in Dutch), NOS (7 December 2021).
- Dutch government resigns over child welfare scandal, Associated Press, via CBS News (15 January 2021).
- Ruim 90 procent gedupeerde ouders klaar met integrale beoordeling (in Dutch), Rijksoverheid (Government of the Netherlands) (21 February 2025).
More accounts to read
All Community Voices-
Families
Debts by Averaging
Australia averaged people's yearly income to decide who had been overpaid welfare. People with irregular pay were chased for debts they did not owe. A Royal Commission called it neither fair nor legal.
2015–2019 · Australia Read the account -
Families
A False Match, an Arrest at Home
Facial recognition wrongly matched a man's licence photo, and police arrested him in front of his daughters. His lawsuit changed how Detroit police may use the technology.
2020–2024 · Detroit, US Read the account -
Students
The Summer an Algorithm Set the Grades
With exams cancelled, a model lowered 39 percent of teacher-assessed A-level grades, and small classes fared best. Students marched; the government reversed within days.
2020 · England Read the account
Summarised from the public record. Follow the sources for the full story.