The use of artificial intelligence in the judicial sphere – new perspectives

Condividi su:

Share on:

Artificial Intelligence: American Cases

Two incidents have recently been reported in the United States of America (in the federal districts of Mississippi and New Jersey), in which federal judges ordered the withdrawal of their own decisions following challenges raised by the parties’ attorneys. In the Mississippi case, the original order contained an erroneous identification of the procedural parties and referred to accusations not included in the initiating documents. In the New Jersey case, instead, the decision contained non-existent case law citations and attributions of statements that did not correspond to the truth, traceable to the unauthorized insertion of research conducted through artificial intelligence tools, subsequently not reviewed. In both cases, the withdrawal order did not formally mention AI as a direct cause, but the specialized press and legal scholarship highlighted the risk of so-called AI “hallucinations,” i.e., the production of fictitious data and references, received without critical review.

Legal Considerations

Constitutional and Procedural Principles

The phenomenon in question must be evaluated in light of key principles of judicial activity: the impartiality and independence of the judge, the subjection of the judge solely to the law, and respect for due process and adversarial proceedings. The use, even if only indirect and non-transparent, of tools that generate factual or legal errors undermines legal certainty and negatively affects the reliability of the judicial function.

Risks Connected to the Use of AI

The use of inadequately controlled artificial intelligence software can give rise to: informational hallucinations (false or invented citations), errors in the representation of facts, arbitrary attributions of precedents and statements. These risks have already been observed on several occasions, to the extent of justifying disciplinary warnings against professionals who have submitted documents based on AI-generated texts without prior verification.

Regulatory Framework

https://www.consilium.europa.eu/it/policies/artificial-intelligence/ Regulation (EU) 2024/AI Act qualifies AI systems intended to support judicial activity as “high-risk,” imposing specific control and recording obligations. The general principle established by the European legislator is that the final decision-making must remain an exclusively human-led activity. Therefore, unauthorized or unsupervised use contradicts the European legislative intent and the principles of proper exercise of the judicial function.

The Doctrinal Perspective of Amedeo Santosuosso

The opinion expressed by Amedeo Santosuosso, Extraordinary Professor at the IUSS Pavia ‘Judging with AI is According to the Constitution,’ deserves mention in a systematic perspective, as it emphasizes the role of the European AI Act as a tool for balanced regulation of AI use in the judicial field. According to this approach, the Regulation does not adopt a technophobic stance but recognizes the possibility of employing AI systems in preparatory, procedural, or integrative tasks, provided that final decision-making remains firmly entrusted to the human judge. AI can contribute to improving the quality of judicial work, provided there is critical review and effective human oversight.

The Opposing Position of Roberto Bichi

In contrast to Professor Santosuosso’s optimistic approach, a recent contribution by Dr. Roberto Bichi, former President of the Court of Milan, outlines a rather critical and cautious position on the use of artificial intelligence in the judicial field. The author distinguishes between weak or medium AI tools (recurring acts, legal research, chatbots for simple acts), considered useful and already widespread, and strong AI systems, aimed at predicting or replacing the judge’s decision, considered incompatible with constitutional principles. The objections are based on intrinsic technological limits, risk of standardization, algorithmic opacity, and compromise of due process. From this perspective, it is necessary to set strict limits on the extent of predictive-decisional AI, avoiding automation paths that would reduce the human judge to a mere executor of algorithmic schemes.

Conclusions

In light of the facts presented and the legal considerations discussed, it can be reasonably stated that: The U.S. cases represent a paradigmatic example of the risks arising from uncontrolled use of AI in the judicial field. The phenomenon of AI “hallucinations,” if not subjected to critical review, is capable of affecting the validity and authority of judicial decisions. In light of the constitutional framework and the European Regulation (AI Act), the use of AI in decision-making must be limited to merely instrumental and preparatory functions, excluding any substitution of the judge’s evaluative activity. It is necessary to adopt strict human oversight protocols both by attorneys and judges, so that the use of technology does not compromise the principles of due process and the protection of fundamental rights. These cases are also highly significant regarding the not always appropriate use of artificial intelligence in the legal sector, highlighting the need to always verify information generated by AI tools and to consult only accredited professionals or when legal intervention is necessary to assess a concrete case. If you have a doubt or need legal advice, contact us https://studiolegalefaraci.ch

Condividi su:

Share on: