Sweden's welfare AI flagged migrants as fraud suspects; privacy watchdog closed the case
Sweden's Social Insurance Agency (Försäkringskassan) used a machine-learning risk score since at least 2013 to select parents claiming temporary parental benefit for fraud investigations. Its own supervisory authority, the ISF, concluded in 2018 that the algorithm did not treat applicants equally, which the agency rejected; a 2024 analysis by Lighthouse Reports and Svenska Dagbladet of more than 6,000 flagged people found it over-selected women, migrants, low-income earners and people without a university education. The data protection authority IMY opened a supervision in June 2025, but after the agency said it had stopped using the risk profile about a month before its 5 September 2025 reply, IMY closed the case on 18 November 2025 without examining it further. In February 2026, under a government assignment, the agency set out a plan to evaluate its existing risk profile in sickness benefit and decide whether to develop more.
the agency's fraud prediction algorithm discriminated against women, migrants, low-income earners and people without a university education
in its current design [the algorithm] does not meet equal treatment
Den behandling som IMY haft för avsikt att granska har därmed upphört.
Försäkringskassan ska även ta ställning till om det finns skäl att utveckla fler riskprofiler
Rejected the ISF's 2018 equal-treatment conclusion; its fraud-algorithm supervisor denied wrongdoing and told reporters he did not think the agency needed to be more transparent.
Closed its supervision because the processing it intended to examine had ceased.
No outcome recorded yet.