AMAS — Austria's Banned-Then-Vindicated Jobseeker Profiling Algorithm

Between 2015 and 2018, Austria's Public Employment Service (Arbeitsmarktservice, AMS) developed a regression-based scoring model, later branded AMAS, that sorted every registered jobseeker into three employability groups (A: high prospects, B: medium, C: low prospects) using variables including age, gender, nationality, care obligations and disability. Group C jobseekers were to be steered away from AMS's more expensive training programmes on cost-efficiency grounds, while Group A jobseekers were also deprioritised on the assumption they would find work unaided.

An independent academic study (Allhutter et al., Frontiers in Big Data, 2020) obtained AMAS's technical documentation and found it explicitly subtracted 0.14 points from a jobseeker's score for being a woman, with a further penalty for declared care obligations applied only to women. Austria's data protection authority banned the system in August 2020 as an insufficiently justified form of automated profiling; AMS deleted the project's data that autumn and, per board chairman Johannes Kopf, wrote off roughly €2.5 million in development costs.

Austria's Federal Administrative Court did not finally resolve the legal challenge until September 2025, ruling that AMAS had not violated GDPR Article 22 because a human caseworker retained final sign-off — but by then AMS had already abandoned the project and confirmed it has no plans to revive it, saying it would likely use different AI-based methods today. AMAS was never deployed nationally and remained a contested pilot for its entire five-year existence.