KOHEN AVOCATS FIRM · PARIS 16
126 Faces Pulled From Jihadist Videos by AI as Paris Jails a Syria Executioner for 30 Years: How France Prosecutes Foreign Fighters

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

126 Faces Pulled From Jihadist Videos by AI as Paris Jails a Syria Executioner for 30 Years: How France Prosecutes Foreign Fighters

Speak to a lawyer 06 89 11 34 45

On 18 September 2026, Interpol announced that a five-day operation run from Tunis had used artificial intelligence to pull 108,076 facial images out of propaganda published by jihadist groups, narrow them to 6,362 usable faces, and identify 126 suspected foreign terrorist fighters so far. The agency says investigative leads from the operation are already supporting court cases in several countries, including one case against two detained suspects, and that thousands of images are still being processed. Four days later, in Paris, the Special Criminal Court sentenced French jihadist Othman Garrido to thirty years’ imprisonment for his role in a public execution in Syria, with twenty years before any parole. The same week, a second Paris terrorism court opened the trial of Sarah Talib, a 31-year-old Frenchwoman accused of joining the Islamic State after travelling to Syria at nineteen. One week, three events, one question: what happens between an AI match on a jihadist video and a French courtroom, and what does French law actually allow?

This article answers that question from the inside of French procedure. France claims one of the widest criminal jurisdictions in Europe over terrorism committed abroad, concentrates every case in Paris in the hands of a specialised prosecution service, tries the gravest charges without a jury, and punishes terrorist murder with up to life imprisonment fenced by a security period that can reach two-thirds of the sentence. Each of those claims has strict legal conditions, and each can be challenged. The reporting below is built on the official Interpol release of 18 September 2026, Le Monde’s coverage of the Garrido verdict of 22 September 2026, Courthouse News’ account of the Talib trial that opened on 23 September 2026, the French Criminal Code and Code of Criminal Procedure in force, and three recent decisions of the Criminal Chamber of the Court of Cassation quoted word for word.

A warning on method: the Talib trial was still running at the time of writing, and she is presumed innocent. The Garrido conviction is a first-instance verdict subject to appeal. Nothing below predicts the outcome of any pending case; it explains the rules of the game.

I. Arrested on Video in Syria, Tried in Paris: How Do French Courts Claim Jurisdiction?

The striking feature of the September 2026 news is geography. The faces were collected from online jihadist content, the suspects may be scattered across a dozen countries, the execution happened in Syria, and yet it is Paris that tries them. That outcome is not an accident of politics. It is the product of a deliberate jurisdictional architecture that France has built over forty years, combining a Criminal Code provision aimed specifically at terrorism abroad, international conventions implemented into domestic procedure, and a specialised Paris prosecution service with nationwide concurrent jurisdiction. Understanding each layer matters, because each layer is also where a defence lawyer looks for a way out.

A. Can France try crimes committed in Syria?

The short answer is yes, in most configurations, and the starting point is an article of the French Criminal Code that was written for exactly this scenario. Article 113-13 of the Criminal Code provides: “La loi pénale française s’applique aux crimes et délits qualifiés d’actes de terrorisme et réprimés par le titre II du livre IV commis à l’étranger par un Français ou par une personne résidant habituellement sur le territoire français.” In plain English: French criminal law applies to terrorist crimes and offences committed abroad by a French national or by a person habitually resident in France. Both Garrido and Talib are French nationals who allegedly joined the Islamic State in Syria, so French jurisdiction over their Syrian conduct follows directly from the nationality prong, without any need for Syria’s consent or cooperation.

The residence prong goes further, and the Court of Cassation confirmed in 2024 just how far. In a case concerning a Palestinian man who had lived in Syria, arrived in France as a refugee in October 2015, and was accused of terrorist conspiracy for acts allegedly committed in Syria, Iraq, Egypt, the Netherlands and Germany between 2014 and 2020, the defence argued that French courts had no jurisdiction over the pre-2015 Syrian conduct because the man had no link with France at the time. The Criminal Chamber rejected that argument on 6 March 2024 (no. 23-87.046, ECLI:FR:CCASS:2024:CR00411). It approved the reasoning that “les dispositions de l’article 113-13 du code pénal s’appliquent aux personnes de nationalité étrangère résidant habituellement en France et n’exigent pas que la fixation de cette résidence préexiste à la commission des faits”, and restated the rule itself: “aux termes de l’article 113-13 du code pénal, la loi pénale française s’applique aux crimes et délits qualifiés d’actes de terrorisme et réprimés par le titre II du livre IV de ce code, commis à l’étranger par un Français ou par une personne résidant habituellement sur le territoire français.” The court added the decisive sentence: “Ce texte ne distingue pas selon que la personne de nationalité étrangère a fixé sa résidence habituelle en France avant ou après la commission, à l’étranger, d’actes de terrorisme.” A foreign fighter who settles in France after the Syrian facts can therefore still be tried in Paris for those facts. For the 126 suspects identified by Operation Shams II, this means any of them who is French or who later lives in France already falls, in principle, within the reach of a French investigating judge.

Beyond the nationality and residence hooks, France has a third jurisdictional layer for terrorism: universal jurisdiction under international conventions. Article 689 of the Code of Criminal Procedure states the general principle: “Les auteurs ou complices d’infractions commises hors du territoire de la République peuvent être poursuivis et jugés par les juridictions françaises soit lorsque, conformément aux dispositions du livre Ier du code pénal ou d’un autre texte législatif, la loi française est applicable, soit lorsqu’une convention internationale ou un acte pris en application du traité instituant les Communautés européennes donne compétence aux juridictions françaises pour connaître de l’infraction.” For terrorism specifically, Article 689-3 of the Code of Criminal Procedure, implementing the 1977 European Convention on the Suppression of Terrorism signed at Strasbourg, allows prosecution in France of, among other offences, wilful attacks on life committed abroad even without any French link, provided the suspect is present in France and extradition is not granted. In practice, prosecutors rarely need this fallback for French returnees, because Article 113-13 of the Criminal Code already covers them, but it matters for the Shams II scenario: a non-French suspect identified by facial recognition, arrested on French soil on a foreign warrant or a red notice, and not extradited, could still face a French courtroom.

None of this makes jurisdiction automatic, and the same 2024 decision shows where the fight happens. The Court of Cassation partly quashed the referral order on a second ground: the lower court had wrongly held that French jurisdiction extended to foreign conspiracy acts merely because they were connected to terrorist acts committed in France, as an inseparable whole. The high court recalled that extension by connexity to a foreign offence against a foreign victim requires a genuine link of indivisibility with an offence committed on French territory, and sent that part back. Defence lawyers in returnee cases therefore attack jurisdiction precisely here: was the accused French at the time, habitually resident, and are the Syrian facts properly qualified as terrorist acts under Title II of Book IV rather than ordinary foreign crimes? Where the file mixes Syrian conduct with later French conduct, they test every claim of indivisibility. Jurisdiction in these files is a battle of characterisation, not a rubber stamp.

One practical consequence follows for anyone identified by an operation like Shams II. The Interpol release stresses that the Baghdad liaison office contributed a large intelligence dataset and that leads are already feeding judicial procedures, including a case with two detained suspects. Intelligence shared between police services does not become court evidence by itself. In France, it must be converted into procedure: a preliminary inquiry or a judicial investigation opened by the anti-terrorism prosecution, formal witness statements, expert facial-comparison reports subject to adversarial debate, and, where coercive measures are needed abroad, mutual legal assistance requests. An AI match is the beginning of a file, not the end of one.

B. Who investigates and which court tries foreign fighters in France?

Every terrorism case in France, wherever the facts occurred, converges on Paris. Article 706-16 of the Code of Criminal Procedure provides that “Les actes de terrorisme incriminés par les articles 421-1 à 421-6 du code pénal, ainsi que les infractions connexes sont poursuivis, instruits et jugés selon les règles du présent code sous réserve des dispositions du présent titre.” Article 706-17 then gives the Paris actors their power: give the Paris anti-terrorism prosecutor, the Paris investigating-judges unit, the Paris criminal court and the Paris assize court concurrent jurisdiction over these offences alongside the ordinary territorial courts The National Anti-Terrorism Prosecution Office, the Paris investigating judges’ unit, the Paris criminal court and the Paris assize court therefore have concurrent jurisdiction over every terrorist offence in the country, including terrorist acts committed abroad where French law applies. A returnee arrested in Marseille, Lille or Lyon is investigated and tried in Paris. The Garrido and Talib trials both sit in Paris for exactly this reason.

The trial court for the gravest charges is not an ordinary assize court. Terrorism felonies go to a specially composed assize court, sitting without a jury. Article 698-6 of the Code of Criminal Procedure sets the composition: “la cour d’assises prévue par les articles 697 et 697-4 est composée d’un président et, lorsqu’elle statue en premier ressort, de quatre assesseurs, ou lorsqu’elle statue en appel, de six assesseurs.” Five professional judges at first instance, seven on appeal: no citizens drawn by lot, a bench of magistrates only. Garrido was tried and sentenced by this formation over seven days of hearings from 14 to 22 September 2026. Supporters of the system argue that professional judges are better equipped to handle complex, document-heavy terrorism files and to resist emotional pressure; critics reply that the absence of a jury removes a democratic safeguard precisely where the stakes, thirty years or life, are highest. Either way, an accused facing the Paris special assize court must understand that the audience deciding guilt is composed entirely of career judges.

The appeal structure was illustrated by the Court of Cassation only months ago. On 28 January 2026 (no. 26-80.097, ECLI:FR:CCASS:2026:CR00263), the Criminal Chamber recorded that “M. [B] [U] a interjeté appel de l’arrêt de la cour d’assises de Paris, spécialement composée, en date du 29 octobre 2025, qui, pour terrorisme et tentative d’importation en contrebande de marchandises prohibées, l’a condamné à dix-sept ans de réclusion criminelle”, with a security period set at two-thirds of the sentence, and, citing expressly “les articles 380-1 à 380-15, 698-6, dernier alinéa, 706-16 et 706-17 du code de procédure pénale”, it ruled: “DESIGNE, pour statuer en appel, la cour d’assises de Paris, spécialement et autrement composée, en matière de terrorisme”. Appeals from the Paris special assize court therefore go to the same court differently composed, and the designation mechanism itself passes through the Court of Cassation. Garrido, convicted at first instance on 22 September 2026, has this appellate path open to him, before professional judges again.

Two features of the investigation phase deserve emphasis for readers following the Shams II identifications. First, French terrorism investigations routinely last years: Garrido spent six years in pre-trial detention before his trial, a duration that reflects the complexity of gathering evidence about Syrian facts, from witness testimony of fellow detainees to authenticated documents to expert analysis of videos and photographs. Facial-recognition matches produced by AI agents, however sophisticated, will face the same adversarial testing as any other expert evidence: the defence can challenge the quality of the source image, the threshold of the match, the chain of custody of the dataset, and whether the “review and verification by trained officers” described by Interpol meets French standards for expert evidence. Second, the financial trail is now part of the standard toolkit. Operation Shams II used cryptocurrency tracing tools that produced three urgent disclosure requests to a virtual-asset service provider, and met gaming-industry representatives about youth radicalisation. In Paris files, terrorist financing charges and asset evidence increasingly sit alongside the core conspiracy charges, because money movements are often easier to prove than presence on a Syrian battlefield.

Finally, where suspects are detained abroad, the mirror procedure is extradition to France, with its own bars and delays. France regularly seeks the surrender of fugitives wanted for terrorism and organised crime, and those requests travel through diplomatic channels with double-criminality checks, specialty protections, and judicial review in the requested state, as illustrated by recent extradition litigation handled through the Versailles appeal court in a Canadian fraud case before the French courts. For Shams II suspects located in cooperative states, a French arrest warrant backed by the Paris investigation can trigger provisional arrest pending the formal request; for suspects in uncooperative zones, the French file waits, and limitation periods for terrorist felonies are long.

II. Thirty Years for a Syria Execution: What Penalties and Defences Apply?

The Garrido sentence stunned even seasoned court watchers: thirty years with a twenty-year security period for a man who, according to Le Monde, did not personally wield the knife but participated in the public execution of two prisoners beheaded by two other French jihadists. The prosecution had asked for life; the defence had begged for a sentence leaving some prospect of release to reward ideological disengagement. To understand the verdict, and to understand what Talib and future Shams II suspects risk, one must read the penalty architecture of French terrorism law: a base conspiracy offence punished as a misdemeanour at ten years, aggravated terrorist murder punished as a felony at up to life, and a security period that can lock the prison door for two-thirds of the term.

A. What sentence do returnees from Syria face in France?

The workhorse charge in returnee files is terrorist criminal conspiracy. Article 421-2-1 of the Criminal Code defines it: “Constitue également un acte de terrorisme le fait de participer à un groupement formé ou à une entente établie en vue de la préparation, caractérisée par un ou plusieurs faits matériels, d’un des actes de terrorisme mentionnés aux articles précédents.” Joining the Islamic State, training, fighting, recruiting, funding, or preparing any listed terrorist act with at least one proven material fact falls within this definition. The punishment is fixed by Article 421-5: “Les actes de terrorisme définis aux articles 421-2-1 et 421-2-2 sont punis de dix ans d’emprisonnement et de 225 000 euros d’amende.” Ten years’ imprisonment and a 225,000-euro fine for participation; thirty years’ criminal imprisonment and a 500,000-euro fine for directing or organising the group. Talib, charged with participating in a terrorist criminal organisation after travelling to Syria in 2014, faces this ten-year framework if the Paris court finds that her Syrian stay amounted to participation characterised by material acts, a question her trial is now testing witness by witness. France 24 also reported on 15 September 2026 that three women of the same family returned from Syria are standing trial in Paris on similar participation charges, which shows how routinely this article is now deployed against female returnees whose roles, according to expert testimony cited at the Talib trial, were long underestimated by investigators who treated women as mere followers.

Where the file contains a killing, the penalty scale jumps to the felony range through the combination of terrorist qualification and homicide. Article 421-1 of the Criminal Code provides that “Constituent des actes de terrorisme, lorsqu’elles sont intentionnellement en relation avec une entreprise individuelle ou collective ayant pour but de troubler gravement l’ordre public par l’intimidation ou la terreur, les infractions suivantes : 1° Les atteintes volontaires à la vie, les atteintes volontaires à l’intégrité de la personne, l’enlèvement et la séquestration ainsi que le détournement d’aéronef, de navire ou de tout autre moyen de transport, définis par le livre II du présent code”. A wilful attack on life committed intentionally in connection with a terrorist enterprise is therefore a terrorist act, and the underlying homicide carries the penalties of Book II. Article 221-1 states: “Le fait de donner volontairement la mort à autrui constitue un meurtre. Il est puni de trente ans de réclusion criminelle.” Thirty years’ criminal imprisonment is the baseline for murder; aggravated forms go to life. Garrido’s conviction for organised murder connected with a terrorist organisation, for his role in the beheading of two prisoners, sits at the very top of this scale, which is why the prosecution could credibly demand life imprisonment and why the court’s thirty-year sentence, though below the maximum, remains one of the heaviest handed to a French Syria returnee.

The figure that will shape Garrido’s life most concretely is not thirty but twenty: the security period. Article 132-23 of the Criminal Code provides: “En cas de condamnation à une peine privative de liberté, non assortie du sursis, dont la durée est égale ou supérieure à dix ans, prononcée pour les infractions spécialement prévues par la loi, le condamné ne peut bénéficier, pendant une période de sûreté, des dispositions concernant la suspension ou le fractionnement de la peine, le placement à l’extérieur, les permissions de sortir, la semi-liberté et la libération conditionnelle.” No suspension, no outside placement, no leave, no day release, no parole during the security period. Its default length is half the sentence, but “La cour d’assises ou le tribunal peut toutefois, par décision spéciale, soit porter ces durées jusqu’aux deux tiers de la peine ou, s’il s’agit d’une condamnation à la réclusion criminelle à perpétuité, jusqu’à vingt-deux ans, soit décider de réduire ces durées.” Twenty years is exactly two-thirds of thirty: the Paris court used the maximum uplift the statute allows, signalling that it placed Garrido’s participation in a public beheading among the gravest conduct short of life. The January 2026 Court of Cassation decision cited above shows the same practice in another terrorism file, where a seventeen-year sentence carried a security period of two-thirds. Readers should grasp what this means in human terms: at thirty-two, Garrido will be fifty-two before he can even apply for conditional release, and any release will then depend on a separate sentence-enforcement judge’s assessment of risk, disengagement and reintegration guarantees.

Two further consequences routinely accompany these sentences. First, ancillary penalties: the January 2026 file combined seventeen years with seven years of socio-judicial supervision, fifteen years of ineligibility and confiscation, a standard package that can include weapons bans, residence bans, and registration in the sex-or-violent-offender files where applicable. Second, civil-party damages: victims and associations routinely join terrorism trials as civil parties, and the Court of Cassation continues to shape that litigation, as in its 19 November 2025 decision on appeals against the civil-party ruling of the specially composed Paris assize court of 4 March 2024. Families of victims of Syrian executions who constitute civil parties in Paris can obtain damages enforceable in France, a remedy that matters precisely because no Syrian court is available.

B. How can the defence challenge an AI-built terrorism file?

The defence in these files does not deny the horror; it attacks the proof, the qualification and the person. Four lines of defence dominate, and each maps onto a weakness of AI-assisted identification files like those emerging from Operation Shams II.

First, contest the identification itself. Interpol states that AI-generated output was reviewed by trained officers under its data-processing rules, and that 6,362 high-quality faces went through its facial-recognition system. A French defence lawyer will demand the underlying material: the original video frames, the extraction scripts, the deduplication parameters, the match scores, the demographic performance of the algorithm, and the qualifications of the reviewing officers. French courts admit scientific evidence only through adversarial expert debate, and a match announced in a press release has no probative value until an expert report is filed, communicated and discussed at the hearing. Where the only evidence placing a suspect in Syria is a blurred frame matched by software, the defence will argue that the standard of proof for a criminal conviction, intimate conviction of the judges, cannot rest on an opaque score. The fact that thousands of Shams II images remain unprocessed cuts both ways: it promises more identifications, but it also shows a pipeline whose error rate has yet to be tested in any courtroom.

Second, contest the qualification. Not every stay in Syria is terrorist conspiracy, and not every presence at a crime scene is murder. Article 421-2-1 requires participation in a group formed or an agreement established with a view to preparing a terrorist act, characterised by one or more material facts. The defence dissects each alleged material fact: was the travel itself the act, or were there training, fighting, guarding, financing or logistical acts? For women returnees especially, the qualification battle is central, as the expert debate at the Talib trial illustrates: the prosecution must prove personal participation, not guilt by marriage or by presence in the caliphate. For Garrido-type murder charges, the defence distinguishes aider, witness and perpetrator, forcing the court to specify each accused’s personal role in the killing. The Le Monde account of Garrido distancing himself from his own past shows this strategy in action: admit the trajectory, dispute the precise criminal role.

Third, humanise the person to contain the sentence. The Garrido trial displayed both faces the court saw: the introspective thirty-two-year-old prisoner of six years and the young fighter of the caliphate. The defence pleaded disengagement, remorse and youth, eighteen at departure, against a prosecution demand for life. It failed to avoid thirty years, but the same arguments routinely move sentences within the scale, decide whether the security period is set at half or two-thirds, and shape the enforcement judge’s later decisions. Mental-health expertise, family guarantees, employment projects and documented deradicalisation efforts are assembled from the first months of detention with this horizon in mind. Talib’s testimony about identity, community and self-esteem, however the court receives it, belongs to the same strategy: explain without excusing, in order to survive the penalty phase.

Fourth, use every procedural remedy, including against the international alert itself. A suspect flagged by a red notice or a diffusion who considers the alert abusive can petition Interpol’s Commission for the Control of Files for access, correction or deletion, in parallel with challenging detention and extradition before the courts of the arresting state. Within France, pre-trial detention is reviewed, investigating judges’ orders are appealed to the Paris investigating chamber, assize verdicts go to the differently composed appeal assize court designated through the Court of Cassation, and final review lies with the Criminal Chamber. The 2024 jurisdiction ruling itself proves that these remedies bite: the referral survived on residence but fell on connexity. A file built on AI identification will offer at least as many attack surfaces as a classic witness-based file, from the legality of data collection to the translation of intelligence into evidence.

Conclusion

September 2026 will be remembered in French counter-terrorism as the month when the future knocked on the courtroom door. Interpol’s Tunis operation showed that the Syrian crime scenes of 2014 are now searchable image databases, that 126 suspects have already been named by software, and that the pipeline is still running. The Paris courts answered the same week in the old language of criminal justice: thirty years and twenty of security for participation in a beheading, a full trial for a woman who left at nineteen. Between the two stands French procedure at its most ambitious, claiming jurisdiction over desert executions through Article 113-13, concentrating every file in Paris through the anti-terrorism prosecution, trying without a jury, and punishing at the top of the scale.

Whether that ambition delivers justice or overreach will be decided file by file, and the tests are already visible. Can an AI face-match survive adversarial expert debate? Can courts individualise guilt inside collective barbarity, distinguishing the fighter, the follower and the witness? Can security periods of twenty years coexist with a genuine policy of disengagement? The coming Shams II prosecutions will force answers, in Paris and elsewhere. Only the judges, weighing each file on its own evidence, can give them.

📄 Official circular