The article examines the “Digital Omnibus” as an act of legislative consolidation aimed at transforming the fragmented structure of European digital law into a coherent normative architecture, thereby laying the foundation for key digital infrastructures, in particular European AI factories. From a doctrinal perspective, it becomes apparent that the previously sectoral regulatory approach (GDPR, DSA, DMA, Data Act, AI Act) has only been able to exert limited steering effect and has, in part, reinforced structural dependencies on non-European technology providers. By contrast, the Digital Omnibus represents an attempt to systematically align data access obligations, interoperability, supervisory competences, and AI regulation, and to consolidate the European Union’s regulatory authority in the digital sphere. The article illustrates how the Omnibus could establish the legal preconditions for AI factories as strategic centres for computing, data, and modelling, and identifies the challenges arising under EU law in this context. From a normative standpoint, it is argued that the Omnibus must be understood as a building block of an emerging digital constitutional order, one that seeks to strike a sustainable balance between freedom of innovation, fundamental rights protection, and public oversight.

The size of the neural networks trained by leading AI companies has sparked extensive debate. Among the issues raised are environmental impacts, training and operational costs, digital sovereignty and compliance, as well as ethical and social concerns. From a scientific perspective, the profoundly flawed theoretical foundations of the technology have been criticised (language models have very little to do with language: Bender and Koller, 2020). From a legal standpoint, the latter aspect is likely the most concerning. If AI systems are to be deployed in areas subject to rules and regulations, they must be capable of encoding data structures that reflect the relevant aspects of the real world. In their current form, however, they are, by design, incapable of doing so. This article examines the use of large language models in a business context. It discusses when and where LLM technology, in its present form, can be applied and identifies areas of concern. Finally, it proposes an alternative in the form of hybrid “small” language models, which combine the strengths of transformer-based architectures and semantic algorithms, thereby also generating additional positive effects in terms of cost, environmental impact, sovereignty, compliance, and ethics.

The paper explores the thesis that, however impressive the performance of artificial in-telligence systems – particularly large language models (LLMs) – their functioning cannot amount to genuine understanding. The analysis moves from philosophical and cognitive accounts of understanding to the mechanisms of symbolic and subsymbolic AI, arguing that the latter exhibit structural correlations with understanding but lack its constitutive dimension: consciousness and normativity. The human mind, though lim-ited by bias, fatigue, and fallibility, possesses self-awareness, experiential depth and normative intentionality that no computational system replicates. Nonetheless, AI out-puts may advance legal reasoning by supporting human understanding, provided that automation bias and uncritical reliance are curbed. The paper concludes with implica-tions for legal education and other institutions, advocating a shift from production to reflexive and ethical engagement with AI.

Generative AI systems are increasingly being used to produce legal academic texts. This applies equally to seminar papers, bachelor’s and master’s theses in legal studies, as well as to publications of all kinds. As part of the Legal AI Project Saarbrücken 2025 (JIPS), conducted in the summer semester of 2025 at Saarland University, legal scholars at different stages of their careers (students, research assistants, and professors) explored how generative AI systems can be effectively employed in the preparation of legal academic work. In this context, both examples of problematic uses of generative AI systems and a range of useful use cases were identified, the latter leading to a significant increase in efficiency in the production of legal texts. The article presents the methodology and the key findings of the project, with a particular focus on student use. The authors conclude that, when applied professionally, generative AI systems provide highly valuable support in the preparation of legal academic work.

This article addresses the challenges, research questions, and initial approaches to solutions for the automation of legal decision-making processes, using the example of the commercial register system and drawing on symbolic artificial intelligence (AI) and machine learning (ML), including natural language processing (NLP). The authors report on their experience from their ongoing research project on “Automated Legal Reasoning”, within which they have been provided by the commissioning entity with the complete set of commercial register data from the Free State of Bavaria. As a concrete case study, the article examines the practically relevant yet legally highly controversial question of whether a usufruct over a limited partnership interest can be registered in the commercial register.

The legal status of machines is explored. A new concept, status virtualis, is a dialectical innovation. The reasoning begins with the thesis being status naturalis, proceeds through the antithesis being status civilis, and culminates in the synthesis being status virtualis. The notion of status virtualis is characterized primarily by different rules of governing in the so-called computer state. The status virtualis situations discussed can be observed in virtual reality, three-dimensional virtual worlds, massive multiplayer online games (MMOGs), mass media, films, and narratives.

AI systems and large language models have increasingly found their way into legal practice in recent years. Legal scholarship has likewise identified numerous potential applications for artificial intelligence. However, it remains unclear which applications of AI legal practitioners consider useful in their daily work and which aspects they regard as important, for example, in assessing the quality of AI-generated texts. To address this gap in the research, this article presents the results of a study in which legal professionals were surveyed regarding their expectations and requirements with respect to AI systems and AI-generated texts.

Many approaches to legal visualisation primarily aim to enable non-experts to achieve a better understanding of legal content than that conveyed by “traditional” legal texts. However, there is still a lack of evaluations confirming their effectiveness. At the same time—pursuing a similar objective—the use of large language models (LLMs) such as ChatGPT has increasingly come into focus. This article contrasts the use of legal visualisations and LLMs with traditional legal texts and empirically examines their respective contributions to laypersons’ understanding of legal content. In doing so, it seeks to provide deeper insight into their utility and to contribute to the development of an empirical foundation.

The research project DECIDE (Data-driven Exploration in Contextual Information on Decisions) aims to enhance the quality of political decision-making processes, improve the efficiency of public services, and strengthen democratic transparency. At its core is the development of a Legislative and Decision Data Space (DS) that links local and regional government decisions with thematically relevant contextual data. Through the structured publication of legislation and decisions as Linked Open Data (LOD), an interoperable, machine-readable data infrastructure is created, enabling data-driven analysis, decision support, and citizen participation. In the pilot cities and regions of Flanders, Ghent, Freiburg, and Bamberg, data-driven applications are being tested—for example, for the analysis of mobility zones, environmental measures, or funding programmes for sustainable construction. These real-world use cases serve to validate the data space concept and to develop transferable models for other European cities and regions. From an academic perspective, DECIDE contributes to the digital transformation of the public sector by integrating concepts from linked data, semantic interoperability, and data governance into municipal decision-making processes. The project combines research, teaching, and practice. Students develop practice-oriented applications and analyses as part of project work, seminars, and theses. The project’s content is incorporated into courses on open data, smart governance, and digital public administration. In addition, study visits to the pilot cities promote exchange with administrative partners and allow students to experience data-driven decision-making processes in real-world contexts. In this way, DECIDE contributes to the training of future professionals and to the sustainable integration of data-driven innovation processes within public administration.

The paper discusses the insights gained from the supervision of software engineering students who are writing programs that detect “dark patterns” on online interfaces. Hence, the supervisors pursue educational goals while introducing students to the field of computer ethics. The detection programs analyze the HTML code of sales portals. The commandments of computer ethics prohibit the causing of harm and require a consideration of the social consequences of the system being designed. Deceptive patterns are harmful to consumers, not to sellers. We hold that the loss of trust in unfair sellers is a more serious problem than just the use of pre-checked boxes. Information technologies become untrustworthy. User consent becomes inefficient because users are burdened with analyzing risks on webpages.

In 2012, the Föderales Informationsmanagement (FIM) was introduced in Germany. It defines which information relating to service descriptions, required data, and procedures for service provision must be collected, made available, and observed in a standardised manner across digitalisation projects nationwide. However, the system did not establish itself automatically. The available materials primarily support specialised system knowledge (methodological expertise) as well as technological specifications. These materials were reviewed with a focus on the knowledge relevant to digitalisation managers and project leads. The result is a body of background knowledge tailored to this target group, along with indications of further steps that would be desirable for them. Ideally, this approach can help to promote the adoption of FIM.

Public administration faces significant challenges as a result of the digitalisation of its services. Using the example of a medium-sized Swiss municipality, this paper demonstrates the potential of applying enterprise architecture management (EAM) methods within the public sector, provided that its specific institutional conditions are taken into account. By means of the modelling language ArchiMate, two viewpoints—Organizational and Technology Usage—are represented and analysed in the form of diagrams. Based on these analyses, and in conjunction with defined requirements, strategic and concrete measures are derived.

End-to-end digitalised citizen services under the Online Access Act (OZG) require a systematic and structured approach to ensure high quality, efficiency, and legal compliance. The findings of the Working Group on Open Design of Digital Administrative Architectures (AG openDVA) regarding the pathway presented in 2024—and further developed since—from statutory text to a digitalised service have resulted in a “production line” for the efficient development and provision of smart citizen services. The process begins with user stories collected within public administration. Using the methodology of Föderales Informationsmanagement (FIM), structured and standardised master data are derived on the basis of the legal framework, covering the administrative service itself, as well as its processes and data fields. Decisions are modelled through rule mapping at the respective process steps of the reference process. Legal compliance is verified immediately following the modelling stage by the competent legislative authority, thereby establishing the basis for implementation. Knowledge graphs are used to describe and interlink this information and to make it available in a machine-interpretable format. Various platforms can search, visualise, and utilise this knowledge to provide applications and, in the future, to extend it further. Following the transformation of knowledge into process and data field information, an initial procedure is created, which is subsequently enriched with additional information. FIM standards and core services such as FIT-Connect ensure interoperability with other (administrative) systems, while supplementary information on data protection and security can further enhance the enrichment of process information. After comprehensive functional and acceptance testing, as well as validation, digitalised citizen services can be deployed and iteratively improved based on monitoring and user feedback. This “production line” for creating OZG-compliant services provides a well-structured, iterative process that spans from initial analysis to continuous optimisation. Its clear structure and emphasis on standards, usability, and legal compliance ensure that services meet the requirements of all stakeholders.

In response to the accelerating uptake of Artificial Intelligence (AI) across the public sector, this concept paper proposes a practical, five-point framework for integrating AI into parliamentary work: strategy, prioritization, training, implementation and governance. Drawing on workshops and early implementations in multiple legislatures, we outline core AI technologies and indicative applications–ranging from text analytics and information extraction to drafting support–while emphasizing privacy, security and responsible use. We discuss organizational prerequisites (quick wins, national-language LLMs and cross-sector support) and adapt a traffic-light deployment model to guide decisions on where cloud, partner-hosted or on-premises LLMs are appropriate. The approach is designed to be locally configurable rather than one-size-fits-all, enabling parliaments to pursue transparent, accountable and cost-effective AI adoption aligned with legal and ethical obligations.

In Germany, discussions have been ongoing for some time on how to modernise civil procedure. Against the backdrop of declining case numbers in civil courts—particularly in disputes involving lower amounts in controversy—the issue of improving access to justice through online proceedings has gained significant attention. Legislation adopted in December 2025 provides for the piloting of such proceedings. Despite conceptual challenges, several elements of the legislation are of particular importance, including the communication platform, the introduction of a procedural document (or basic document), and the incorporation of the regulatory sandbox concept (Reallabor).

The growing cooperation between humans and machines is reshaping the way society interacts and all sectors are being affected by this transformation, including the Judiciary. Within this context of profound change, the research seeks to critically assess the tangible benefits and emerging risks of human–machine cooperation within the Brazilian Judiciary. Specifically, it aims to: examine the trajectory of the Judiciary’s digital transformation; identify artificial intelligence systems currently under development, as well as those already implemented or in operation within the courts; and analyze the potential implications of human–machine cooperation for the paradigm of justice. Methodologically, the study adopts a deductive approach, combining qualitative and quantitative perspectives and relying on bibliographic and documentary research techniques. The findings indicate that the Brazilian Judiciary has been undergoing an official process of digital transformation since 2006. The analysis on the adoption of artificial intelligence systems reveals that several courts have either developed their own solutions or are in the process of doing so. Moreover, the data show that most machine-assisted tasks are currently confined to administrative functions, repetitive procedures, and features designed to support human activity. More recently, research and development efforts have begun to focus on decision-making systems. From the analysis of the ideal of justice, the study identifies potential benefits, but also emerging risks such as excessive objectivity in decision-making, loss of contextual and semantic interpretation, the standardization and massification of judgments, and technological power asymmetries. The article concludes by proposing regulatory and ethical guidelines aimed at safeguarding fundamental rights and preventing the transformation of justice into an opaque, automated decision-making system devoid of human connection.

The presented article seeks to demonstrate the state of eJustice across selected European countries, most notably in relation to their use of advanced technological solutions, specifically automation, within the judicial process. The presented countries demonstrate varied level of not only automation and the possibilities of automation in the near future, but more crucially they display great variety on the „lower“ technological level of a mere electronisation and digitalization of judicial procedures. By describing these differences and current status via analysis of national reports this article seeks to show that while advanced automation presents an interesting and emerging opportunity for eJustice, the current fragmented approach and the state of infrastructure across Europe does not provide ample opportunity for its realization.

Against the background of developing a digital base or procedural document, this article presents a pilot study examining how generative artificial intelligence can be used to automatically generate relation tables from written submissions in civil proceedings. To this end, a limited number of available pairs of submissions (claimant’s and defendant’s pleadings) are used as test material in order to generate relation tables automatically by applying Retrieval-Augmented Generation (RAG) and Cache-Augmented Generation (CAG). In addition, the study explores whether generative AI can also be used to reconstruct pleadings from published judgments (reverse document engineering), which could in turn serve as training material for relation generation.