The Danish government deployed algorithms for its centralized tax collection system EFI (short for Et Fælles Inddrivelsessystem or one shared collection system) (Algorithm Watch 2019). The system was initiated in 2005 in order to build a digital tax collection system at both the local and national level. However, due to critical structural and technological flaws, the system lost billions of Danish kroner for the public in the form of expired or uncollected claims (ibid). Legal expert Hanne Marie Motzfeldt attributes the failure to mistakes in EFI’s “data, design, programming and integration in the administrative bodies led to administration in conflict with the law” (ibid). Moreover, she argues that “precise knowledge about the functioning of data and business processes that were ‘cast’ into the IT systems were largely placed with the IT provider” (ibid).
The EFI demonstrates the problematic and complicated nature of public administration’s over-reliance on private firms to develop AI software. Private firms, operating on a different incentive-structure, lack the same public oversight that helps guarantee high standards of transparency and respect fundamental rights such as privacy and data protection.
