KOHEN AVOCATS FIRM · PARIS 16
Italy Would Not Send Carla Zambelli Back to Brazil: Two Cassation Rulings on Fair Trial and Prison Assurances, and What a Paris Court Would Do

Hassan KOHEN acts in urgent matters, from the police station to the assize court. Initial strategic analysis: €80 incl. VAT, personal reply within 24 hours.

100% confidential · Legal professional privilege · No obligation

Article generated by artificial intelligence, through a process designed and supervised by the firm

Source: French Court of Cassation – Judilibre and Légifrance open data

Paris Bar Police custody, investigation, assize court CNB listing avocat.fr
Maître Hassan KOHEN, avocat au Barreau de Paris
Hassan KOHEN
Attorney at the Paris Bar

Italy Would Not Send Carla Zambelli Back to Brazil: Two Cassation Rulings on Fair Trial and Prison Assurances, and What a Paris Court Would Do

Speak to a lawyer 06 89 11 34 45

In September 2026, international media reported that Italy had blocked the extradition of Carla Zambelli, the former Brazilian federal deputy and close ally of Jair Bolsonaro, sentenced in Brasilia to a total of more than fifteen years in prison across two cases. Behind that headline lie not one but two finely reasoned judgments of the Italian Court of Cassation, and they deserve a careful reading by anyone facing — or defending — an extradition request in Europe. In the first, judgment no. 21634 of 22 May 2026, published on 11 June 2026, the Sixth Criminal Section annulled without remand the Rome Court of Appeal’s decision authorising surrender for a ten-year sentence imposed for hacking the computer system of Brazil’s National Council of Justice: the conviction, the Court held, was tainted by a serious objective doubt about the impartiality of the trial judge. In the second, judgment no. 27350 of 21 July 2026, the same Court quashed the appeal decision authorising surrender for a separate five-year-and-three-month sentence for carrying a firearm and coercing a man during the 2022 presidential campaign — but this time it sent the case back to Rome for a fresh assessment, because Brazil’s diplomatic assurances did not show with sufficient precision that the designated women’s prison could meet Zambelli’s medical needs.

For an international audience, and in particular for anyone who could be arrested in France on a foreign warrant, these two rulings are a masterclass in how European extradition judges actually reason: they do not retry the facts, they do not declare a foreign justice system globally unfair, and they do not accept diplomatic assurances at face value. They test, point by point, whether surrender in this individual case would expose this person to a flagrant denial of justice, to inhuman detention conditions, or to a punishment contrary to French and European public policy. This article explains both Italian decisions in detail, then answers the question our clients ask first: what would a Paris court do with the same file?

I. Two Brazilian Convictions, Two Extradition Requests, Two Different Answers From Rome

A. Why Italy refused surrender for the ten-year hacking sentence: the judge who combined every role

Carla Zambelli, elected federal deputy for São Paulo and one of the most visible supporters of former president Jair Bolsonaro, was convicted by the Brazilian Supreme Federal Court for directing the hacker Walter Delgatti’s intrusion into the electronic system of the National Council of Justice (Conselho Nacional de Justiça, CNJ) between 2022 and 2023. The intruders manipulated official documents — certificates, release orders, banking-secrecy authorisations and arrest warrants — including a fabricated arrest warrant against Supreme Court Justice Alexandre de Moraes himself. Prosecutors presented the operation as a deliberate attack on the justice system, and the Supreme Federal Court unanimously sentenced Zambelli to ten years’ imprisonment for hacking and ideological falsehood. After her conviction she left Brazil — press reports trace a route through Argentina and the United States before she reached Italy, where she also holds citizenship by descent — and presented herself publicly as a political exile. Brazil issued an arrest warrant, an Interpol alert followed, and the Italian police eventually arrested her in Rome.

The Rome Court of Appeal, ruling in March 2026, authorised her extradition subject to conditions drawn from Brazilian assurances: detention in a specific women’s prison in the Federal District, access for Italian diplomatic and consular authorities, and periodic reporting on her health and detention conditions. The Court of Cassation took a different view, and its reasoning is what makes this judgment important beyond Brazil and Italy. The decisive figure was Justice Alexandre de Moraes. He was not merely one judge among others on the convicting panel: he was the reporting judge (relator), he took part in deciding the objection raised against his own participation, he then issued the arrest warrant and was involved in the extradition request — while also being, potentially, a victim of the very conduct prosecuted, since a fake arrest warrant had been forged against him and inserted into the judicial database. The Italian Court did not state that Justice de Moraes was personally biased. It held something more structural and, for defence lawyers, more powerful: even where a judge believes himself impartial, the proceedings must also appear impartial to an outside observer, and here that objective appearance of impartiality was missing. Nor was it enough to answer that Brazilian law provides remedies in theory or that the conviction was issued by a collegiate court. What mattered was whether Zambelli’s objection to the judge’s participation had been examined by a genuinely separate and neutral authority — and the Court found that this had not been demonstrated. It therefore annulled the Rome decision without remand, ordered her release unless she was held for other reasons, and she walked out of prison. The judgment acquits no one and declares no general finding that Brazil cannot guarantee fair trials; it says that this particular conviction, produced by this particular combination of roles, cannot support extradition from Italy.

Three clarifications matter for readers tempted to draw sweeping conclusions. First, Zambelli’s Italian citizenship was not the ground of refusal: the Court expressly decided on impartiality, not nationality. Second, the political context of the case — a Bolsonaro ally convicted by a court that is itself a central actor in Brazil’s institutional confrontation — was not treated as making the offences political offences; the refusal rested on procedural fairness, not on politics. Third, the ruling leaves the door open, in principle, to a differently constituted prosecution: the defect identified attaches to the specific proceedings and the specific combination of functions, not to the Brazilian Supreme Federal Court as an institution.

B. Why the five-year firearms sentence went back to Rome: assurances must prove medical fitness, not just name a prison

The second Brazilian request concerns an entirely different conviction. On the eve of the 2022 presidential run-off, in São Paulo, Zambelli pursued a man while armed — an episode captured on video and widely broadcast — and she was ultimately sentenced to five years and three months’ imprisonment for unlawful carrying of a firearm and coercion with a weapon. The Rome Court of Appeal authorised extradition for this conviction too, again subject to conditions: detention at the Penitenciária Feminina do Distrito Federal, known as Colmeia, unrestricted access to her lawyers and to Italian diplomatic and consular representatives, and periodic Italian monitoring of her health and detention conditions. The defence raised the full arsenal — political persecution, lack of judicial impartiality, breach of the natural-judge principle, double criminality, prison conditions and incompatibility of detention with her state of health — and the Court of Cassation, in judgment no. 27350 of 21 July 2026, rejected all of them except one.

On impartiality, the second judgment deliberately confines the first. In the firearms proceedings the rapporteur was Justice Gilmar Mendes, who also signed the arrest warrant; Justice de Moraes sat on the panel and voted for conviction, but he was not a person injured by these offences and did not cumulate the same pervasive combination of functions, and the case had been allocated through the ordinary electronic system. The structural concern decisive in the hacking case was therefore absent, and the Court made the point explicit: the first decision was never a general finding that any panel including Justice de Moraes is incapable of a fair trial. The Court likewise rejected the argument that trial at first and only instance before the Supreme Federal Court violates the right to appeal: the Supreme Court’s original jurisdiction over sitting members of Congress has a constitutional basis, and Article 2 of Protocol No. 7 to the European Convention on Human Rights expressly allows an exception to review by a higher tribunal where the person is tried at first instance by the highest court — a conclusion the Cassation had already upheld when Italy extradited Henrique Pizzolato, the former Banco do Brasil director, to Brazil in the Mensalão corruption affair (Court of Cassation, Section 6, judgment no. 10965 of 11 February 2015). The political-offence argument failed under the Italy-Brazil extradition treaty, which identifies a political offence primarily by the legal interest affected: offences against personal liberty, individual safety and public security do not become political merely because the convicted person is a politician or because the prosecution unfolds in a heated political climate.

The one ground that succeeded is the one defence teams across Europe should study most closely. The Court accepted that the general conditions at Colmeia did not in themselves expose Zambelli to a real risk of inhuman or degrading treatment — recalling its recent line, including Section VI judgment no. 20942 of 3 June 2026, that Brazilian prisons are no longer treated as a matter of common knowledge automatically barring surrender, and that the defence must adduce specific, current and reliable material showing a real individual risk. But general conditions were not the issue. The issue was Zambelli’s documented medical needs, and on that point Brazil’s assurances, however detailed about the building, the visits and the reporting, did not demonstrate with sufficient precision that this prison could actually provide the care she requires. The Court therefore quashed the appeal decision and remitted the case to a differently composed Rome Court of Appeal for a fresh assessment limited to that medical question. The lesson is sharp: assurances are not judged on their solemnity but on their specificity, their reliability and their verifiability against the individual file — and a generic designation of a compliant prison, plus promises of monitoring, will not cure a gap on medical fitness. As of late September 2026, that remitted assessment is the live front in the Zambelli saga, and it is the part of the story most directly transferable to French practice.

II. What a Paris Court Would Do With the Same File: the French Filters and the Defence Playbook

A. How French judges test a foreign conviction: nationality, political purpose, double criminality and the tribunal that guarantees fundamental rights

Suppose Brazil had found Zambelli in Paris rather than Rome and requested her surrender from France. The first thing a French lawyer explains to the client is that extradition in France is a judicialised, multi-stage procedure governed by the Code of Criminal Procedure, and that each stage is a genuine filter — not a rubber stamp. The starting rule is set by Article 696-2 of the Code of Criminal Procedure: “Le gouvernement français peut remettre, sur leur demande, aux gouvernements étrangers, toute personne n’ayant pas la nationalité française qui, étant l’objet d’une poursuite intentée au nom de l’Etat requérant ou d’une condamnation prononcée par ses tribunaux, est trouvée sur le territoire de la République.” (The French government may surrender, at their request, to foreign governments any person without French nationality who is the subject of proceedings brought in the name of the requesting State or of a conviction pronounced by its courts and who is found on the territory of the Republic.) Nationality is therefore the threshold question. Here the Zambelli facts produce an immediate Parisian answer: she holds Italian citizenship, not French citizenship, so the absolute bar of Article 696-4, item 1 — “L’extradition n’est pas accordée : 1° Lorsque la personne réclamée a la nationalité française, cette dernière étant appréciée à l’époque de l’infraction pour laquelle l’extradition est requise” (Extradition is not granted where the requested person has French nationality, assessed at the time of the offence for which extradition is requested) — would not protect her in France. A French passport saves you from extradition; a European Union passport alone does not.

The same Article 696-4 then lists the substantive bars that would do the real work in a Zambelli-type file, and each maps directly onto one of the Italian debates. Item 2 refuses surrender “Lorsque le crime ou le délit à un caractère politique ou lorsqu’il résulte des circonstances que l’extradition est demandée dans un but politique” (where the crime or offence is of a political character, or where the circumstances show that extradition is sought for a political purpose). French courts interpret this narrowly, much as Rome did: holding office, belonging to an opposition movement or being prosecuted during an institutional crisis does not by itself make hacking a judicial database or chasing a citizen with a firearm a political offence, but a file showing selective prosecution of opponents, exceptional procedures or a requesting judge pursuing a personal conflict can feed the parallel argument under item 7 — “Lorsque la personne réclamée serait jugée dans l’Etat requérant par un tribunal n’assurant pas les garanties fondamentales de procédure et de protection des droits de la défense” (where the requested person would be tried in the requesting State by a court that does not ensure the fundamental guarantees of procedure and of defence rights). It is under this item 7 that a Paris chambre de l’instruction would conduct exactly the exercise the Italian Court performed: not judging whether Justice de Moraes is biased in his heart, but examining from the outside whether the cumulation of roles — injured party, rapporteur, judge of his own recusal, signatory of the arrest warrant, actor in the extradition request — preserves the objective appearance of impartiality that French and European law demand. Item 6 adds a further French shield with no exact Italian twin in this case: refusal where the requesting State’s penalty is contrary to French public policy (“Lorsque le fait à raison duquel l’extradition a été demandée est puni par la législation de l’Etat requérant d’une peine ou d’une mesure de sûreté contraire à l’ordre public français”), the provision under which Paris polices death-penalty exposure, irreducible life terms and penalties that shock the conscience of French law.

Two further technical filters complete the Parisian examination, and both would have mattered for Zambelli. Article 696-3 sets the penalty thresholds for extraditable offences, beginning with “Tous les faits punis de peines criminelles par la loi de l’Etat requérant” (all acts punishable by criminal penalties under the law of the requesting State) and extending to correctional offences carrying at least two years’ imprisonment or, for a convicted person, a pronounced sentence meeting the treaty minimum — thresholds that ten-year and five-year sentences clear without discussion, which is precisely why the real battle in heavy-sentence files always moves to bars and guarantees rather than thresholds. Double criminality then requires that the conduct also be punishable under French law: the CNJ hacking translates naturally into Article 323-1 of the Criminal Code — “Le fait d’accéder ou de se maintenir, frauduleusement, dans tout ou partie d’un système de traitement automatisé de données est puni de trois ans d’emprisonnement et de 100 000 € d’amende” (fraudulently accessing or remaining in all or part of an automated data-processing system is punishable by three years’ imprisonment and a 100,000 euro fine), with five years where data were altered — and the armed pursuit into French provisions on violence, threats and weapons offences. Finally, Article 696-6 protects the surrendered person after the event through the specialty rule: “Sous réserve des exceptions prévues à l’article 696-34, l’extradition n’est accordée qu’à la condition que la personne extradée ne sera ni poursuivie, ni condamnée pour une infraction autre que celle ayant motivé l’extradition et antérieure à la remise” (subject to the exceptions in Article 696-34, extradition is granted only on condition that the extradited person will be neither prosecuted nor convicted for an offence other than the one that motivated the extradition and predating the surrender) — a guarantee that matters acutely where, as in Zambelli’s situation, two convictions and two requests coexist and the requesting State might be tempted to use one surrender to serve the other file.

B. The Paris procedure step by step: investigating chamber, Court of Cassation, government decree — and where the defence wins release

Procedure is where French extradition law gives the defence its sharpest tools, and where the Italian rulings find their closest Parisian echo. A person arrested in France on a foreign warrant is brought promptly before the public prosecutor, and if she declares that she does not consent to surrender, Article 696-15 of the Code of Criminal Procedure seizes the investigating chamber (chambre de l’instruction) of the court of appeal, which hears the person within ten working days and delivers a reasoned opinion within a month: “Cet avis est défavorable si la cour estime que les conditions légales ne sont pas remplies ou qu’il y a une erreur évidente” (the opinion is unfavourable where the court considers that the legal conditions are not met or that there is a manifest error). The same article strictly channels review: “Le pourvoi formé contre un avis de la chambre de l’instruction ne peut être fondé que sur des vices de forme de nature à priver cet avis des conditions essentielles de son existence légale” (an appeal against the investigating chamber’s opinion may be based only on formal defects capable of depriving the opinion of the essential conditions of its legal existence). If the opinion is favourable, surrender is then authorised, outside the consent case, by executive act — Article 696-18 provides that “l’extradition est autorisée par décret du Premier ministre pris sur le rapport du ministre de la justice” (extradition is authorised by decree of the Prime Minister on the report of the Minister of Justice) — itself challengeable before the administrative courts. And throughout, liberty is litigable: Article 696-19 states that “La mise en liberté peut être demandée à tout moment à la chambre de l’instruction selon les formes prévues aux articles 148-6 et 148-7” (release may be requested at any time from the investigating chamber in the forms provided for in Articles 148-6 and 148-7), with short time limits and the possibility of supervision measures — the provision under which a Zambelli-type detainee in Paris would systematically seek release while the substantive arguments mature.

The French Court of Cassation has built, case by case, exactly the kind of demanding review Rome has just applied. On fair trial, the landmark is the Criminal Chamber’s judgment of 28 March 2023 (no. 22-84.382, ECLI:FR:CCASS:2023:CR00286), given on an Italian request for the surrender of a man sentenced to life imprisonment by the Milan assize court of appeal: the Court recalled that “Il se déduit de l’article 6 de la Convention européenne des droits de l’homme relatif au droit à un procès équitable qu’un Etat partie à la Convention, requis aux fins d’extradition d’une personne, a l’obligation de s’assurer que celle-ci ne sera pas exposée, dans l’Etat requérant, à un déni de justice flagrant pouvant résulter, notamment, de l’impossibilité d’obtenir qu’une juridiction statue à nouveau sur le bien-fondé de l’accusation, alors qu’elle a fait l’objet d’une condamnation in abstentia au sens de la jurisprudence de la Cour européenne des droits de l’homme” (it follows from Article 6 of the European Convention on Human Rights on the right to a fair trial that a State party to the Convention, requested to extradite a person, is obliged to ensure that she will not be exposed, in the requesting State, to a flagrant denial of justice that may result in particular from the impossibility of obtaining a fresh determination of the merits of the accusation, where she has been convicted in absentia within the meaning of the European Court of Human Rights’ case law). The investigating chamber, seised of such a complaint, must verify that the convicted person, if she so wishes, will be able “d’obtenir qu’une juridiction statue à nouveau, après l’avoir entendue, sur le bien-fondé de l’accusation en fait comme en droit” (to obtain a fresh ruling by a court, after hearing her, on the merits of the accusation in fact and in law) — and if the requesting State’s procedure fails that test, the chamber issues an unfavourable opinion. Transposed to the Zambelli hacking file, this is the French doctrinal home of Rome’s impartiality holding: a Paris chamber would demand proof that the objection to the judge’s cumulated roles was examined by a genuinely separate authority, and that a retrial or equivalent review cures the structural defect — failing which, unfavourable opinion, exactly as in Rome.

On assurances, the leading Parisian authority is the Criminal Chamber’s judgment of 30 January 2024 (no. 23-83.549, ECLI:FR:CCASS:2024:CR00067), given on a Tunisian request concerning murder charges carrying a potential death sentence. The Court quashed a favourable opinion because the Tunisian note verbale was ambiguous — calling the guarantee request premature, invoking a thirty-year moratorium and international commitments rather than a clear undertaking — and it stated the governing principle without nuance: “Le premier de ces textes prohibe l’extradition vers un autre État d’une personne dont il y a des motifs sérieux et avérés de croire qu’elle serait exposée dans le pays de destination à un risque réel d’être soumise à la peine de mort” (the first of these texts prohibits extradition to another State of a person where there are serious and substantiated grounds to believe that she would face a real risk of being subjected to the death penalty in the destination country). The method is identical to Rome’s in the second Zambelli judgment: assurances must be specific, issued by an authority capable of binding the State, and purged of ambiguity — and a Paris chamber seized of a medical-fitness file would apply the same test to prison-health guarantees, ordering, where needed, a supplementary information measure (complément d’information) and refusing to close the file until the requesting State proves, document by document, that the designated facility can deliver the required care. The most recent confirmation of this strictness came on 27 May 2026 (no. 25-84.904, ECLI:FR:CCASS:2026:CR00687), where the Criminal Chamber reviewed an Algerian request for enforcement of a life sentence imposed in absentia and upheld the investigating chamber’s granular, partially favourable approach — proof that Paris disaggregates files offence by offence and guarantee by guarantee, just as Rome split Zambelli’s two requests instead of ruling once for everything.

The practical consequence for anyone arrested in France on a foreign warrant is therefore clear, and it is the same lesson Rome teaches. Do not expect the court to declare the requesting country’s justice globally unfair; expect it to test your individual file against the bars of Article 696-4, the specialty shield of Article 696-6, the fair-trial review distilled from Article 6 of the Convention, and the assurances case law on death penalty, inhuman treatment and now medical fitness. Assemble from day one the specific, current and reliable material the courts demand: the exact procedural history showing who judged whom and who decided the recusal, the prison’s real medical capacity demonstrated by records rather than brochures, the prior record of the requesting State in honouring its undertakings, and the treaty and statutory texts in their current version. Request release early and repeatedly under Article 696-19, seek a supplementary information order to force the requesting State onto the record, and prepare the appeal and the administrative challenge to the surrender decree in parallel. That is how a Zambelli-type file is defended in Paris — and, as Italy has just shown twice in three months, that is how such files are won in Europe.

Conclusion

The Zambelli rulings will be cited well beyond the Brazil-Italy corridor because they state, with unusual clarity, the modern European grammar of extradition. A ten-year sentence for attacking a justice system’s computers cannot be enforced abroad where the conviction was produced by a judge cumulating the roles of victim, rapporteur, judge of his own challenge, arresting authority and extradition actor — objective impartiality is not a courtesy but a condition of surrender. A five-year sentence for an armed assault on the eve of an election can survive every classical objection — political offence, natural judge, double criminality, general prison conditions — and still stall on a single precise question: can this prison treat this patient, proved on paper, verifiably, now. And a Paris court, applying Articles 696-2 to 696-19 of the Code of Criminal Procedure, Article 323-1 of the Criminal Code, Article 6 of the European Convention and the Court of Cassation’s exacting case law on flagrant denial of justice and diplomatic assurances, would reason in exactly the same way. For requested persons and their counsel, the message of September 2026 is therefore double-edged but hopeful: European judges will not shelter you behind your passport or your politics, but they will hold the requesting State to its burden of proof — on the fairness of the trial that convicted you and on the humanity of the cell that awaits you — and where that proof fails, they will do what Rome did: refuse, or send the file back until the State proves its case.

📄 Official circular