Berkes RudolfOn March 5, 2025, the GENEZIS Excellence Student Workshop of Pázmány Péter Catholic University, in collaboration with the Department of Environmental Law at PPKE JÁK and the Institute of Legal Studies at HUN-REN Centre for Social Sciences, organized a panel discussion examining the environmental and constitutional law issues related to the application of artificial intelligence (AI). The speakers agreed that, similar to environmental protection, the regulation of AI applications also adopts a human-centered approach, thus there are fundamental directions and ethical principles that are considered universal worldwide.
Participants of the Roundtable Discussion:
- Ágnes Tahyné Kovács, Associate Professor, Senior Research Fellow (PPKE JÁK, TVKI)
- Boldizsár Szentgáli-Tóth, Senior Research Fellow (HUN-REN TK JTI)
- Ferenc Petruska, Lieutenant Colonel, Associate Professor, and Head of Department (NKE-HHK Department of Defense Law and Administration)
- Gábor Kecskés, Associate Professor, Research Fellow (SZE ÁJK, HUN-REN TK JTI)
Ágnes Tahyné Kovács provided a working definition of AI as technology trained on large amounts of data, capable of communicating with the external world, not only performing targeted inferences but also making other inferences. However, the traceability and interpretability of these processes are complicated by the well-known 'black box' problem associated with AI's opaque operation. From a legal perspective, the expectation is to make this process transparent. She divided the issue of environmental protection into four parts: the effects on the individual (1) and society (2), as well as the environmental burden from software development (3) and hardware application (4). Regarding environmental protection, she noted that AI is not inherently good or bad. While AI has a potentially high environmental footprint, it can also enhance the efficiency of numerous tasks and research from an environmental perspective.
Boldizsár Szentgáli-Tóth, leader of the Algorithmic Constitutionality Momentum Research Group, pointed out that their research project is oriented towards the assessment of fundamental rights in the context of the expanding social role of AI, determining the circumstances in which risks arise and those in which benefits are realized. The foundation of algorithmic constitutionality therefore, lies in the enforcement of fundamental rights during the use and regulation of AI, thereby exemplifying the human-centered approach previously mentioned by Gábor Kecskés. He further emphasized that the rights to a fair trial, freedom of speech, and a healthy environment are key areas of focus for the research group.
From a constitutional law perspective, there are three main regulatory directions to consider. The first, and perhaps the most widespread, is the adoption of non-binding soft law, such as ethical codes, which establish the principles for AI use. The second direction involves the establishment of normative law, otherwise known as hard law, which is enacted through the conventional legislative process. An exemplar of this is the EU AI Act, which was adopted last year and governs AI applications by classifying risks and establishing requirements and sanctions for developers and users.
The complexity of the phenomenon to be regulated, coupled with lawmakers' reluctance to over-regulate and stifle innovation, hampers regulation. As normative law significantly lags behind technological development, courts often bear the burden of applying existing law to complex phenomena like AI use. However, judicial application is still in its infancy, necessitating the development of tests for courts. The primary objective of the research group is to propose judicial tests by the end of the project that comprehensively interpret statutory requirements and constitutional principles in AI-related cases.
Gábor Kecskés believes that environmental law, largely covered by soft law, could serve as a useful analogy for AI regulation. The historical underpinnings of environmental protection can be traced back to the fundamental goal of safeguarding human well-being. This notion finds historical precedent in the early days of international environmental law, which emerged in 1902 with the agreement among participating states to protect birds deemed beneficial to agriculture. Despite divergent state regulations, the principles of environmental law have the potential to foster unity through a human-centered approach. A similar observation can be made about AI regulation, which is also widely regarded as an anthropocentric framework, prioritizing human safety, freedom, and societal organization, albeit with distinct emphases across American, European, and Chinese systems. Due to the transnational nature of the environmental challenges, this has led to a situation in which many states seek to regulate environmental protection within their borders, with the result that the law is only enforceable and applicable in a very limited geographical area. The diverse manifestations of adverse impacts give rise to varied state responses. In certain countries, concerns over survival, for instance, due to rising sea levels, dominate state actions, while in others, the pursuit of opportunities becomes the primary concern. This diversity of objectives and challenges hinders the establishment of consensus, thereby diminishing the efficacy of normative law. In this context, Kecskés proposes that the precautionary principle, employed in environmental law, offers a transferable legal framework for AI regulation. This principle posits that the optimal course of action for the environment, even when considering the deployment of AI technology, is to select the solution that is based on the best available scientific knowledge.
Ágnes Tahyné Kovács emphasized the principle of cooperation, noting that due to the rapid pace of development, it is no longer sufficient to review best practices every three to five years; it must be done almost weekly. Quoting Pope Francis, she suggested that "algorethics" could serve as a bridge between developers, manufacturers, users, and decision-makers, providing a framework for cooperation.
Boldizsár Szentgáli-Tóth sees a shift towards a more constructive phase in scientific discourse after years of pessimism, making it particularly important for representatives of different scientific fields to engage in dialogue, considering the complexity of the problem, and strive for solutions through an interdisciplinary approach.
Ferenc Petruska highlighted that AI is a meta-juridical phenomenon. AI possesses cognitive capabilities that make legal regulation challenging. It can act as a multiplier, which humans cannot keep up with. He outlined four characteristics with examples that make humans uncompetitive: autonomy, speed, endurance, and scalability. For instance, AI fundamentally undermines the application of copyright law, as if we cannot determine who wrote or drew a work, we cannot establish the author's identity either. The objective is to develop an appropriate validation strategy, as it remains difficult to regulate technical details, so Petruska suggests a more effect-based approach.
He anticipates a gradual and measured adoption of technology in society, tempered by prevailing concerns, primarily the lack of specialized knowledge necessary to utilize AI effectively. An illustrative example of this distrust is that while we easily tolerate errors from human colleagues at rates over 10%, even a 3-5% error rate from AI can hinder its adoption. Petruska further posits that, despite the advancements in technology, there will be domains in which human contributors will continue to add value. Focus should be on developing fundamentally human skills alongside mastering AI use.
Gábor Kecskés looking into the future, emphasized that AI will not replace lawyers; rather it will provide a competitive advantage to those who employ it proficiently. Preserving expertise is crucial to effectively interpret the results, as the utilization of AI is contingent upon the system's reliability and the capacity to provide valuable insights. If this is not possible, it is better to forgo AI use.
Boldizsár Szentgáli-Tóth also highlighted the importance of validation, as checking AI's work remains a critical human capability. Recognizing errors, mistakes, and frauds is a fundamental human aptitude, however, its cultivation is imperative to address the emergent risks stemming from AI's expanding social role.
__________________________________________________________
This report was prepared with the support of the Algorithmic Constitutionalism Research Group (LP2024-20/2024), funded by the Hungarian Academy of Sciences.
__________________________________________________________
The views expressed above belong to the author and do not necessarily represent the views of the Centre for Social Sciences.