Analysis of AI Security Requirements in Public Procurement Tenders

Public procurement tenders provide insights into the security requirements organizations expect from AI systems. As governments and public institutions increasingly adopt AI-based solutions, their tenders may specify requirements related to AI security, robustness, data protection, and resistance to adversarial attacks. However, these requirements are often embedded in lengthy and heterogeneous tender documents, making it difficult to systematically assess which security aspects are addressed, how they are formulated, and how they vary across sectors and use cases.

The goal of this thesis topic is to develop an automated crawling and analysis pipeline to collect public procurement tenders and extract AI security requirements from their documents. Through an empirical analysis of the collected data, the thesis aims to identify recurring security requirements, characterize their distribution, and investigate how AI security expectations evolve across public institutions and application domains.