This article examines the relationship between Artificial Intelligence and administrative discretion in public administration. It argues that the widespread assumption that discretionary decisions cannot be automated rests on an overly simplified, binary understanding of discretionary power. Drawing on doctrinal analysis and administrative practice, the article demonstrates that discretion operates in different degrees and modalities and is often structured by guidelines, methodologies, and established administrative practice. In many cases, formally broad discretionary powers function in practice as what the article describes as “pseudo-discretion”, where standardised criteria significantly constrain the range of possible outcomes. The article, therefore, suggests that the key issue in the debate is not whether Artificial Intelligence may exercise discretion as such, but rather what type of discretion is involved. While decisions requiring genuine individualised judgement may remain inherently human tasks, forms of structured or pseudo-discretion may, in principle, be capable of being replicated through automated systems.

The regulation of energy innovation increasingly tests legal frameworks’ ability to adapt to unanticipated innovations. Nuclear law, where complex and safety-oriented regulatory frameworks combine with increased interest in novel technology like small modular reactors (SMRs), is a good case study for this tendency. SMRs are touted as a way to combine nuclear energy's reliability with modern energy systems' flexibility, scalability, and adaptability. At the same time, their implementation calls into question the legal frameworks, which were created for huge, traditional nuclear power facilities. The Czech Republic provides a valuable case study of this broader regulatory dilemma. This article examines whether exemption-based regulations can serve as an adequate and sustainable approach to governing SMRs and exposes the tensions that arise when a system of case-by-case exemptions is used as a substitute for detailed regulation.

The increasing automation of administrative punishment within smart cities marks a transformative shift in public law enforcement. While automated decision-making enhances efficiency and consistency, it raises critical concerns about its true incentives and adherence to the principles of administrative punishment. This article explores the mechanisms and legislative frameworks enabling smart administrative punishment, focusing on traffic enforcement and the fiscal incentives that shape policy. It critically examines the risks of revenue-driven enforcement, erosion of legal safeguards, and distortion of administrative priorities. By analyzing cases such as the objectivization of vehicle owner liability in the Czech Republic, the article shows how poorly regulated automation may compromise legal principles like proportionality, due process, and deterrence. It argues that while efficiency is a legitimate goal, economic rationality should not overshadow justice, calling for a balanced approach that integrates automation with safeguards against misuse.