Electronic devices are fundamental building blocks for today’s world. As a result, it is essential to verify the security and establish trust not only in software but also in the underlying hardware components. Hardware security covers a wide range of research and engineering domains that create methods or primitives to achieve this goal. Among others, this includes Side-Channel Analysis (SCA), reverse engineering, and security primitives such as Physical Unclonable Functions (PUFs). The development and availability of powerful Artificial Intelligence (AI) methods becomes a powerful complement to traditional statistical algorithms to analyze the security of systems. For example, classic Machine Learning (ML) methods like Support Vector Machines (SVMs) or linear discriminant analysis (LDA) have been extensively researched, successfully applied, and integrated into the Common Criteria (CC) certification process of security chips. ML methods, AI, and Convolutional Neural Networks (CNNs) also facilitate to learn structures in complex PUF circuits to assess their security. In recent years CNNs are increasingly being employed in SCA and image recognition during netlist reverse engineering. The growing popularity of such Neural Network (NN)-based approaches is accompanied by a computationally costly and thus economically demanding hyperparameter search. This can be a concern in certification contexts when budgets are limited and a limitless search is not a viable solution. On a more general scale, AI creates the issue of how to define the greatest possible security level that eliminates as many assumption and estimation errors as possible.