Using AI against repeat offenders: the solution to the problem or a danger?
Judges and lawyers analyze for Nació the proposal by the Barcelona Prosecutor’s Office to create an AI-based calculator to deal with habitual thieves—an option already being studied by the Department of Justice.
Is artificial intelligence an effective tool against repeat offending? Could it streamline legal procedures with guarantees and reduce court overload—which fosters repeat crimes and a sense of impunity? The chief prosecutor of Barcelona, Neus Pujal, has put forward the idea of implementing an “electronic repeat-offense calculator” powered by AI to quickly determine whether a detainee meets the criteria for imprisonment.
The tool would check if the person has at least three firm convictions for thefts exceeding a total of €400—and above all, if they are record‑able. If so, they could be charged with a lesser crime, which is penal. The aim is to assist prosecutors and improve the fight against habitual thieves, and sources from the Department of Justice told Nació that they are already working on it without a concrete deadline. Meanwhile—and pending further details that the Barcelona Prosecutor’s Office has not provided to this paper—various legal professionals have expressed doubts and reservations.
“Not everything is AI and AI is not magic”
Jurisdiction Professor Jordi Nieva (UB) and Civil Law Professor Mireia Artigot Golobardes (UPF), who works on AI application in law, both question whether this calculator really counts as artificial intelligence. “Today we call everything AI,” says Artigot, noting that if it’s simply a systematized calculation tool based on a database, it’s automation—and then the debate is no different than whether one should use Microsoft Office. “The question here would be why hasn’t a system like this existed for some time?” Nieva suggests.
The key issue is whether it includes an AI that learns from cases and evolves how it makes calculations or makes inferences to reach results—which contradicts legal principles and poses risks to defendants’ fundamental rights. “In criminal law, caution must be maximal, because we are working with custodial sentences and irreversible implications,” Artigot emphasizes, working on AI‑law projects.
In this regard, Carlos Pascual, coordinator of Judges for Democracy (JJpD) in Catalonia, explains that similar calculators are already used in other areas of Spanish justice—and while they are not determinative (final decisions are always made by judges), they are not sufficiently questioned. This includes the controversial VioGén algorithm—Spain’s integrated gender-violence monitoring system used to assess victim risk to inform police protection levels. Also the Veripol tool, used by Spain’s National Police to detect false complaints, which was withdrawn in March due to lack of validity in judicial procedures.
Artigot argues that pressure to find solutions for repeat offending amid case accumulation shouldn’t lead to cosmetic measures without a prior strategic framework. She calls for protocols and mechanisms ensuring full guarantees and validation systems before launching such tools. “We all know something must be done, but we talk about AI as if it were magic. We cannot build the house from the roof down,” she states.
Judges and lawyers’ requirements
There is broad consensus across parliamentary lines identifying court overload due to repeat-offender cases as one of the judicial system’s main problems and a major security challenge. “We’ve grown accustomed to court responses coming late, which is an anomaly,” Nieva remarks. The General Council of the Judiciary (CGPJ) has agreed on the priority opening of 34 new courts in Catalonia in 2026, which will be added to five others agreed by the State, Generalitat and Barcelona City Council, while legal professionals continue demanding solutions. However, they also emphasize requirements and red lines regarding the Prosecutor’s proposed calculator.
Judges for Democracy emphasize result oversight, requiring that the algorithm be transparent, auditable and reviewable in order to avoid experiences like VioGén and Veripol. Likewise, the Barcelona Bar Association (ICAB) demands guarantees that the tool will respect defendants’ rights, while recognizing positively any proposal that improves the functioning of the justice administration.
Similarly, Artigot requests specifics regarding the corpus of data the AI would use, warning of personal data protection risks. She also demands detailed information on how the artificial intelligence would be trained, how it would integrate into the judicial system’s operation, how quality and fundamental rights defense would be guaranteed, and who would be responsible. Above all, how it would be ensured that a human decision remains final. “Simply validating the AI result is not a human decision,” she notes.
AI’s fit within the judicial system
Justice is one of the “high‑risk” AI use areas under EU legislation. That means it’s permitted—but not in the “unacceptable risk” category—and its use will be strictly conditioned unless regulatory changes occur. Nevertheless, Nieva highlights three primary uses already possible: using AI to draft judgments; to correct biases in information transmitted in a trial; and to predict the risk of repeat offending when applying remand decisions.
These limitations mean that Judges for Democracy align with Nieva in foreseeing only limited impact from the prosecutors’ proposed repeat‑offense calculator in reducing case resolution times. “The judge will still need to deliberate, and many cases won’t be AI‑intervened,” they say, reiterating that the only real solution is adding more courts and staff. “If trials were faster, fewer cases would arrive because the sense of impunity would end,” Nieva concludes. All of this awaits whether the Government ultimately decides to implement this electronic repeat‑offender calculator.
Source: NacióDigital, Fer servir la IA contra la multireincidència: la solució al problema o un perill?, published 26 July 2025