Who is Accountable to Citizens? AI, Security, and the Erosion of Accountability Mechanisms in Democratic Systems

Anna Sroka San Pablo CEU University (Madrid, Spain) anna.sroka@ceu.es The integration of artificial intelligence within security decision-making processes, including threat analysis, resource prioritisation, risk assessment and operational support, is precipitating a paradigm shift in the logic of accountability within democratic systems. The present paper addresses the key research question of who is accountable to citizens when public decisions are co-produced by algorithmic models and the underlying assumptions are probabilistic, opaque and challenging to debate publicly. The research thesis posits the hypothesis that the proliferation of AI within the security sector may result in the erosion of democratic accountability mechanisms. This erosion is theorised to occur through several mechanisms, including: the blurring of the decision-making chain; the shift of responsibility to 'technical recommendations'; the increase in dependence on technology providers; and the weakening of parliamentary and judicial control due to the limited auditability of systems. In the following analytical section, a framework is proposed for assessing the "accountability deficit" based on three criteria: In order to ensure the validity of the decision-making process, it is imperative to establish the chain of accountability, that is to say, to identify the parties who approved, supervised and configured the system. Furthermore, it is essential to ensure that the decisions made are both explainable and appealable, that is to say, that there is a clear rationale for the decisions taken and that citizens have a genuine avenue for appeal if they feel that their concerns have not been addressed. Finally, the effectiveness of sanctions and corrective measures must be evaluated in order to ascertain whether errors result in institutional consequences and meaningful system improvements.