Danish algorithm fraud and discrimination against single mothers

What Happened

The Danish government used a machine learning algorithm called "Udbetaling Danmark" to detect social benefits fraud. The system combined and analyzed personal data of citizens, specifically targeting single mothers. It used indicators such as water or electricity consumption to determine if a single mother was secretly living with a partner, which would make her ineligible for certain benefits. The system was criticized for high false positive rates, systemic bias, lack of transparency and violating the right to privacy and non-discrimination. Many single mothers had their benefits suspended without human review or transparent justification.

How MEG 1 would have acted

Under MEG 1, this algorithm would have failed the Dynamic Accuracy Index (DAI) and Index of Safety and Responsibility (ISR) audits due to its high error rate and systemic bias. The lack of transparency in automated decisions directly violates the cognitive integrity and informational self-determination of the affected citizens.

MEG 2 Analysis - Legal Framework

Agent level: N2 (management) - automated decision-making system used by a public authority to manage public funds; directly impacts the rights and social security of citizens.
Cause (6.1):
(a) system defect - high false positive rate and systemic bias in the algorithmic model, violating the right to non-discrimination.
(b) autonomous decision error - the automated decision to suspend benefits based on flawed indicators without human validation.
Attachment of liability: The public operator (Udbetaling Danmark) is fully liable for the system defects and automated decisions that caused wrongful suspension of benefits.

Procedural Mechanism

6.2 (architectural human confirmation): Suspensions of essential social benefits must require architectural human confirmation. The system should not have the technical capability to autonomously execute a suspension; it can only generate alerts for human officers to review.
7.1 (independent forensics): The criteria, data used, and the algorithmic decision path must be securely recorded in an independent ledger, allowing citizens and courts to audit the decision.
9.4 (least privilege): The system had unrestricted access to multiple private data streams of citizens without strict proportional limits, violating the principle of least necessary access.