Outsourcing artificial intelligence in public administration

Abstract

For the planning and execution of public services (and occasionally even public functions in the strict sense), outsourcing is an established approach. This also holds true for the processes of digitizing infrastructure, auxiliary services, and computing resources that are functional or otherwise used to support the performance of public duties. In this regard, the issue of the algorithm’s provider acting as an extraneus (in a subjective sense) to the administration is a recurring one. The concern presented at the starting point of the study is whether standard interpretative categories remain relevant when the goal of outsourcing is an artificial intelligence-based product or solution, or does artificial intelligence outsourcing alter the discourse’s terms? The analysis shows how, despite the validity of some interpretive categories, the AI importing in the context of public powers presents unique characteristics that stem from the market structure in which AI is produced, marketed, and tested as well as from certain implicit, functional, characteristics. As such, these characteristics should be emphasized. Specific issues that occur in the operation of outsourcing artificial intelligence services in the context of public authorities can be verified using the AI Act and the GDPR.