Ádám EszterOn June 25, 2026, the Institute for Legal Studies hosted the sixth session of its ongoing event series, Sustainable Use of Artificial Intelligence in Legal Work. Launched in the fall of 2025 in collaboration with the National Laboratory for Artificial Intelligence and the Lendület/Momentum Research Group on Algorithmic Constitutionalism, the series serves as a specialized forum for evaluating the intersection of advanced technology and legal frameworks.
The primary focus of this session, titled Recent Lessons and Ongoing Challenges in the Research of Algorithmic Constitutionalism, was the formal presentation of the nascent ConstAI database. Developed under the aegis of the Lendület/Momentum Research Group, the ConstAI repository compiles global normative legislation, soft law instruments, and judicial precedents concerning the algorithmic impact on three core fundamental rights: freedom of expression, the right to a fair trial, and the right to a healthy environment.
The first session of the proceeding featured presentations by Boldizsár Szentgáli-Tóth - senior research fellow of ELTE CSS Institute for Legal Studies, principal investigator of the Research Group on Algorithmic Constitutionalism; Rudolf Berkes - researcher, ELTE CSS Institute for Legal Studies; and Dóra Pálfi – doctoral candidate, Károli Gáspár University of the Reformed Church Faculty of Law, who introduced the database's methodology, structural organization, and preliminary comparative data across distinct subfields. The subsequent segment of the event transitioned into a formal manuscript discussion of a forthcoming study titled The implementation of algorithmic constitutionalism in the field of sustainability and planetary well-being, co-authored by Boldizsár Szentgáli-Tóth, Rudolf Berkes, and Orsolya Tuba. Lilla Király, representing the Károli Gáspár University of the Reformed Church, served as the official opponent, delivering a comprehensive academic critique of the manuscript's theoretical and empirical assertions.
The discussion addressed the multifaceted intersections of sustainable artificial intelligence, fundamental rights, and constitutional law. The proceedings commenced with an exposition by Boldizsár Szentgáli-Tóth, who delineated the interim research outputs of the Algorithmic Constitutionality Research Group. The primary objective of this project is the systematic codification and comparative analysis of global normative frameworks, soft law instruments, and judicial precedents concerning artificial intelligence. The research scope is meticulously restricted to three core constitutional pillars where a proliferation of novel legal disputes is anticipated: freedom of expression, the right to a fair trial, and the right to a healthy environment. To facilitate this academic inquiry, the research group is developing a unique, open-access, English-language repository designated as the “Const AI” database. Guided by a rigorous, methodologically justified codebook, the database disaggregates complex legislative acts – such as the European Union’s Artificial Intelligence Act – into distinct, provision-level data units based on their fundamental rights implications. Conversely, soft law entries, including international conventions promulgated by the Council of Europe, are indexed via their overarching principles, while judicial decisions are coded by synthesizing both their operative clauses and ratio decidendi. Szentgáli-Tóth emphasized that given the highly uniform societal and institutional challenges precipitated by global digital platforms, comparative constitutional law serves as an indispensable tool for identifying transnational regulatory trends ahead of the project’s scheduled conclusion in 2029.
The discourse subsequent shifted to the contemporary landscape of artificial intelligence regulation within the United States, presented by Rudolf Berkes, who elucidated a profound federal regulatory vacuum. At present, a comprehensive federal statutory framework governing artificial intelligence remains non-existent, with the legislative output of Congress restricted to highly localized measures such as the Take It Down Act, which strictly addresses the non-consensual proliferation of AI-generated pornographic deepfakes. Deep-seated political polarization and institutional gridlock within the federal legislature have impeded the passage of broader, bipartisan regulatory proposals. Consequently, a pronounced fragmentation of the legal environment has occurred, driven by a surge in subnational legislative activity. The research database currently documents over three hundred enacted state-level statutes, which exhibit a clear normative consensus primarily regarding the prohibition of obscene synthetic media and the mitigation of algorithmic interference in democratic electoral processes. Certain jurisdictions, notably California, New York, and Illinois, have progressively sought to impose heightened compliance, auditing, and disclosure obligations upon developers of frontier AI models. In stark contrast, the federal executive branch has pursued an aggressively deregulatory stance designed to safeguard American technological hegemony by incentivizing infrastructure expansion and energy grid optimization. This divergence has precipitated an acute constitutional conflict concerning regulatory jurisdiction, wherein federal attempts to curtail state-level initiatives via statutory moratoria or restrictive executive orders have faced significant pushback from the Senate. This jurisdictional tension is further compounded by national security considerations, exemplified by the recent invocation of stringent federal export controls to restrict global access to advanced proprietary models, thereby compelling tech developers to institute rigorous identity and citizenship verification protocols for ordinary users.
The judicial and procedural ramifications of generative artificial intelligence were evaluated by Dóra Pálfi, who charted the evolution of judicial standing orders and bar association ethics opinions over the preceding eighteen months. Following early, highly publicized instances of professional misconduct involving algorithmic "hallucinations" and fabricated legal citations, the initial judicial response was characterized by absolute procedural prohibitions. However, this prohibitive stance has gradually transitioned toward a disclosure-based paradigm, wherein litigants are required to submit formal declarations detailing the specific models utilized and the human verification mechanisms employed. Current professional trends prioritize the human-in-the-loop doctrine, emphasizing that practitioners possess a non-delegable ethical duty to independently verify algorithmically retrieved information. Pálfi demonstrated that even specialized, proprietary legal databases are susceptible to generating spurious data, as evidenced by recent transnational case law where built-in AI tools hallucinated non-existent judicial precedents. Furthermore, the integration of automation disrupts traditional hourly billing structures and raises unresolved questions regarding the equitable allocation of initial software procurement costs. Within the judiciary itself, the deployment of AI raises profound constitutional concerns regarding the obligation to provide authentic, human-authored judicial reasoning. While tribunals have occasionally utilized generative tools to expedite administrative workflows, superior constitutional courts have strictly admonished that outsourcing the drafting of the ratio decidendi to an algorithm unconstitutionally hollows out the judicial function. Moreover, public administrations face an influx of voluminous, AI-generated petitions that are structurally complex yet substantively deficient, forcing an institutional reliance on automated screening mechanisms. This reciprocal automation risks engendering automation bias, a psychological phenomenon wherein human adjudicators progressively defer to algorithmic determinations without critical independent reflection, thereby threatening the core tenets of due process.
The final segment of the proceeding comprised a manuscript debate for a forthcoming study co-authored by Boldizsár Szentgáli-Tóth, Rudolf Berkes and Orsolya Tuba, which theorizes a conceptual nexus between algorithmic constitutionality and the doctrine of planetary wellbeing.
Tuba argued that planetary wellbeing constitutes an ecocentric framework that directly challenges contemporary tech narratives centered exclusively on human utility and economic productivity, positing instead that human rights are structurally interdependent with the ecological integrity of the biosphere. The study constructs a triadic conceptual framework to evaluate this intersection, commencing at the foundational level with the formal constitutional recognition of environmental rights. The second level establishes an algorithmic analogy, mapping the tenets of algorithmic constitutionality onto environmental law by demonstrating that multinational corporate polluters and digital platform conglomerates operate identically as opaque, unaccountable private power centers capable of encroaching upon fundamental rights without democratic legitimacy. The tertiary level outlines a model of collective action, vowing to distribute transparent and enforceable obligations across the state, platform operators, industrial actors, and the citizenry.
Lilla Király provided a comprehensive academic critique of the manuscript, advocating for a calibrated realignment of the text to alter its current composition of ninety percent theory and ten percent practice to a more pragmatic sixty-forty ratio. To substantiate the urgency of the environmental dimensions of algorithmic governance, Király introduced critical empirical metrics, noting the massive quantities of clean water consumed during model training phases and the significant carbon footprint associated with routine generative prompts. She highlighted that major tech conglomerates have reported substantial increases in aggregate emissions driven by data center infrastructure, prompting speculative investments into extraterrestrial and deep-ocean data storage solutions. Király strongly urged the integration of a consumer protection dimension, reframing the individual not merely as a passive subject of rights, but as an active, informed consumer empowered by novel regulatory instruments such as the European Union’s directives on the green transition and the Digital Product Passport.
Furthermore, she emphasized the necessity for absolute terminological precision within the manuscript, stating that the analysis must rigorously distinguish between standard foundation models and the exponential proliferation of corporate AI agents. While foundation models are numerically limited, autonomous AI agents operate continuously within closed corporate frameworks to optimize internal data workflows in compliance with data protection regulations, yet their cumulative, uninterrupted operational cycles generate an ecological footprint that far exceeds that of conventional generative interfaces. The co-authors acknowledged these critiques, committing to integrating these consumer-centric frameworks, empirical realities, and precise technological typologies prior to finalizing the manuscript for peer-reviewed journal submission.
_________________________________________________________
This report was prepared with the support of the ELTE CSS Algorithmic Constitutionalism Lendület/Momentum Research Group (LP2024-20/2024), funded by the Hungarian Academy of Sciences.
_________________________________________________________
The views expressed above belong to the speakers and do not necessarily represent the views of the Centre for Social Sciences.