IntelliGuide Logo

Artificial Intelligence and Court Decisions
a Lawyer’s Perspective

WHEN THE ALGORITHM TAKES THE JUDGE’S SEAT

What happens when Artificial Intelligence enters a courtroom? The implications are profound and touch every aspect of the legal profession. In this article we explore together the risks and opportunities of a justice system confronting the machines, without surrendering its soul.

( by: Paolo Guerra | date: 20/04/2026 )
Il processo nell’era dell’Intelligenza Artificiale.

The legal process in the age of Artificial Intelligence.

There is a moment in every lawyer’s professional life when the gaze lifts from the case files and turns toward the future, and toward the prospects of this ancient profession. And what I glimpse today, listening to the debate on Artificial Intelligence applied to law, deeply unsettles me. I picture a courtroom in 2050: no gown, no closing argument and, above all, no human involvement with the defendant whatsoever. Just a terminal loaded with codes and an algorithm that, in a flash, analyzes legal precedents and ultimately delivers ‘statistically probable’ verdicts.
All of this will undoubtedly be efficient and objective, but I ask myself: will it still be justice?

Platone, le ombre digitali e il diritto che dovrebbe essere.

Plato, digital shadows and the law that ought to be.

Twenty-four centuries ago, in his Republic, Plato described prisoners chained in a cave, forced to watch shadows projected onto the wall, convinced that those silhouettes were reality. Only those who managed to break free were able to step into the light and contemplate the Truth of things. In the age of Artificial Intelligence, we risk something very similar: precisely at the moment when we believe we have achieved the highest rationality through technology, this could metaphorically lead us back into the cave theorized by the Greek philosopher.

Today, however, the shadows would no longer be cast by fire, but by a computer screen.
The algorithm, by its very nature, works on patterns and statistical correlations — in practice, on what has already been (the shadows). But the law, the real one, whose purpose is to repair wrongs and save lives, should rightly aspire to constant improvement, to what ‘ought to be’ and what, hopefully, will be (the truth).
This distinction, which philosophers define as the leap from ‘being’ to ‘ought to be’, is something no neural network can, and most likely will ever, truly understand. A boundary that no technological progress will ever fully manage to erase, and that justice can never afford to ignore.

L’avvocato ‘in carne ed ossa’: una presenza da difendere.

The ‘real-life’ lawyer: a presence worth defending.

In the system taking shape in the wake of Artificial Intelligence, the lawyer is increasingly becoming an appendage, a ‘bottleneck’, if not outright an ‘avoidable cost’.
In this regard, allow me to recall what this legal professional truly does. First and foremost, he listens, translating his client’s concerns into legal language. He then finds, within the folds of the law, room for the human element, for nuance and exception, for those particular circumstances that no dataset, due to its intrinsic ‘rigidity’, could ever contemplate.
Hannah Arendt, a twentieth-century German philosopher and political theorist, spoke of the ‘banality of evil’: the possibility of committing atrocities while simply following the rules, through the suppression of critical judgment.

At this point, it is legitimate to ask: would a fully automated judicial system not risk ‘institutionalizing’ something similar, in the legal sphere? The tendency would be toward technically correct but morally blind verdicts, delivered without anyone pausing to ‘truly think’ about them.
And yet, that ‘someone’ whose very function is to reflect and question, ultimately assuming responsibility for decisions, has had a precise name for a very long time: he’s called ‘lawyer’, of course.

Un rischio concreto dell’uso dell’IA: l’inversione dell’onere della prova.

A concrete risk of AI: the reversal of the burden of proof.

There is a specter haunting the supposed ‘algorithmic future’ of law: the risk of a progressive reversal of the burden of proof.
It’s indeed very likely that, in a judicial system in which Artificial Intelligence were entrusted with the task of predicting the probability of guilt based on historical, social and behavioral data, it would no longer be necessary to prove an defendant’s guilt: it would be the defendant himself who would have to prove his own innocence.
This would effectively mark the end of the presumption of innocence, a cornerstone of Western law since the times of Cesare Beccaria, father of modern criminal law.

Goya titled one of his celebrated etchings ‘El sueño de la razón produce monstruos’, meaning ‘the sleep of reason produces monsters’, pointing to the evident danger of abandoning reflection.
Perhaps, considering the future of law, we should update this warning: for an ‘excess of reason’ (in this case, computational), could also, potentially, produce monsters, especially when one forgets that justice is not merely a ‘matter of optimization’.
In conclusion, it must be stressed that a defendant is not, and must never be, a mere ‘statistical data point’: no algorithm will ever be able to replace human sensitivity in judging them.

COMPAS: quando l’algoritmo decide della libertà di una persona.

COMPAS: when the algorithm decides a person’s freedom.

In the United States, a system known as COMPAS, an acronym for Correctional Offender Management Profiling for Alternative Sanctions, has existed for years, and is used across numerous states to calculate the probability of recidivism and guide decisions on bail, parole and sentence length.
In 2016, a ProPublica investigation analyzed the data of over seven thousand defendants in Broward County, Florida, finding that COMPAS assigned African American defendants a recidivism risk almost twice as high as that of white defendants, even in cases where they had not gone on to commit any new offense (*1).

The case of Eric Loomis, sentenced to eight and a half years in Wisconsin partly on the basis of his supposed ‘score’, was emblematic of the flaws of such a ‘futuristic’ procedure: his lawyer was never permitted to examine the internal logic of the computer system that had, in substance, contributed to determining his client’s fate.
This demonstrates the substantial danger that, in the age of AI, justice may be entrusted to opaque and uncontestable mechanisms, removed from any form of verification, yet capable of impacting decisively on the freedom of human beings. A reality that is already with us, and that the law cannot continue to ignore.

Note:
*1: Angwin J., Larson J., Mattu S., Kirchner L., Machine Bias, ProPublica, May 23, 2016.

I benefici dell’IA nella sentenza: un alleato, ma non un giudice.

The benefits of AI in sentencing: an ally, but not a judge.

It would be intellectually dishonest to ignore the concrete benefits that, beyond the possible risks, Artificial Intelligence can bring to judicial decision-making.
Tools based on this technology are already capable of analyzing thousands of legal precedents in a matter of seconds, supporting judges in identifying those most relevant to the case at hand and contributing to greater uniformity of treatment between similar cases. Other systems are able to automatically flag procedural inconsistencies before these compromise the outcome of a trial, or to rapidly anonymize judicial documents, easing the bureaucratic burden weighing on courts.

It’s no coincidence that the European AI Act of 2024, while classifying AI systems applied to justice as ‘high-risk’ tools, explicitly permits their use in a support capacity for judges (*1).
A distinction that the Italian legislator has incorporated with equal clarity in Law 132 of 23 September 2025, whose Article 15 expressly reserves to the magistrate every decision on the interpretation of the law, the evaluation of facts and evidence, and the adoption of measures (*2).

Notes:
*1: EU Regulation 2024/1689 (AI Act), Annex III, point 8;
*2: Law of 23 September 2025, n. 132, art. 15, paragraph 1;

Un futuro da scrivere, non da subire.

A future to write, not to endure.

This article does not intend to induce panic, but rather to represent a call to awareness. I say this because the future of ‘digital justice’ is far from written, let alone predetermined. Its path will need to be traced on the basis of conscious choices. Choices concerning which systems to adopt and what limits to impose on them, guided by the values we consider, case by case, non-negotiable.
The Greek philosopher Aristotle distinguished between ‘technē’, technique, the ‘knowing how to do’, and ‘phronesis’, practical wisdom, the ‘knowing how to judge’. Artificial Intelligence should be considered pure technē. Justice, on the other hand, requires phronesis: context, judgment and, above all, humanity.
On the basis of these premises, we will be able, and indeed must, use this new technology to serve justice, without ever allowing it to replace it.

The images on this page were created using generative Artificial Intelligence tools.