AI-based behaviour analysis in prisons: A multidisciplinary assessment of an Austrian case study

The present study addresses current developments in the field of AI-based surveillance to prevent violence and suicide in prisons. In doing so, we draw on research activities and pilot tests recently conducted in Austrian detention facilities. After providing an overview on the current state of art, the specific technical approach of the project is explained, and its implementation in selected penal institutions is described. The AI system and its purpose of application is then subjected to thorough analysis. We address relevant provisions of data protection law and specific domestic requirements of the penal system as well as the regulation of the technology by the European AI Act. Furthermore, the practical necessity and suitability of the surveillance system are assessed, and key ethical positions regarding its future use are discussed. The entire case study is framed by criminological considerations on the usefulness of deploying such AI-based surveillance technologies. Our examination shows that although the automated prevention of violence and suicide would be ethically justifiable, there is currently no empirical evidence for its suitability. In addition, the analysis reveals that there is no appropriate legal basis for its live application. Moreover, we identified significant shortcomings in the risk classification framework of the European AI Act. Finally, it becomes evident that the development of AI-based surveillance to automatically identify and manage risks in the penal system is largely detached from considerations of actual effectiveness. In consequence, less intrusive measures such as mental health care or improving spatial conditions tend to be neglected in favour of technological innovation. The study concludes with a summary focussing on the identification of regulatory deficiencies and further recommendations for the criminal justice system.