On 10 September 2026, the Supreme Court of India refused to let 1993 Mumbai serial blasts convict Abu Salem walk free. A Bench of Justices Vikram Nath and Sandeep Mehta dismissed his plea for release, holding that the 25-year cap on his detention flows from a sovereign assurance given to Portugal and does not turn his life sentences into a fixed 25-year term. Because Salem was first arrested on 11 November 2005, the courts treat the 25-year period as expiring only in November 2030, and the Court rejected his attempt to backdate the clock to his 2002 detention in Portugal or to deduct remission. For an international audience the ruling is worth reading closely, because it shows how an extradition promise made by one government to another can bind the sentencing state for decades, shape every remission calculation, and survive repeated challenges. France operates on the same logic, with its own strict tests. This analysis explains what New Delhi promised Lisbon, why the early-release claim failed, and what a Paris court would examine before surrendering anyone on comparable assurances.
I. Why does Abu Salem stay in jail until 2030 after being extradited from Portugal?
The short answer is that Salem was not surrendered unconditionally. Portugal handed him over in 2005 only after India gave a formal sovereign assurance, and the Supreme Court has enforced that assurance exactly as written: no death penalty, no detention beyond 25 years, but also no discount beyond what the assurance says. Every argument for an earlier release has now been tested and rejected, first by the Bombay High Court on 15 April and then by the Supreme Court, which found no infirmity in that reasoning.
A. What promise did India give Portugal to obtain the surrender?
Abu Salem, a former member of the Dawood Ibrahim syndicate, was detained in Portugal in 2002 after Interpol issued a Red Corner Notice at India’s request, and he was extradited from Portugal to India in 2005 in connection with the Mumbai blasts case. The extradition only happened because the Indian government gave a sovereign assurance that he would neither be sentenced to death nor imprisoned for more than 25 years. The assurance dates back to December 2002, when the then Deputy Prime Minister L. K. Advani assured the Portuguese authorities that Salem would not face the death penalty or imprisonment for a term beyond 25 years, and the Supreme Court has repeatedly treated that commitment as binding on the Indian state.
The context explains why Lisbon insisted on such a guarantee. In the blasts case, Salem was found guilty of transporting weapons from Gujarat to Mumbai for use in the serial bombings, which killed 257 people and left more than 700 seriously injured. After his return he was sentenced to life imprisonment by special courts under the Terrorist and Disruptive Activities (Prevention) Act in 2015 for the 1995 murder of businessman Pradeep Jain and in 2017 for his role in the 1993 Mumbai serial blasts. Under ordinary Indian sentencing law those life terms could have kept him in prison far longer, so without a cap Portugal would not have surrendered him. The assurance therefore functioned as the legal bridge between a requesting state that retains life imprisonment and a requested state that refuses to extradite without a ceiling. In July 2022 the Supreme Court commuted the life sentences imposed on Salem in the two cases to a maximum term of 25 years in view of that sovereign assurance, while declining to convert the life sentences into fixed-term sentences of 25 years as a matter of sentencing law.
This distinction between a cap on execution and a new sentence is the key to the whole file. The Court held in July 2022 that the Centre was bound by the sovereign assurance and would have to advise the President to exercise the power of remission once Salem completed 25 years in prison, but it left the underlying life sentences intact. Anyone reading the September 2026 ruling must keep that architecture in mind: the convictions remain life sentences, affirmed by the competent courts, and the 25-year limit operates only as an outer boundary imposed by the promise to Portugal. That is why the arithmetic of the release date matters so much, and why the Court refused to let defence counsel restart the clock from an earlier point or shorten it with remission credits.
French extradition law starts from the same premise that surrender can depend on promises. Article 696 of the Code of criminal procedure provides that, absent an international convention providing otherwise, the conditions, procedure and effects of extradition are determined by the provisions of that chapter, which also apply to points not regulated by conventions. Article 696-2 of the Code of criminal procedure then states the basic rule in these terms: “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.” A Paris extradition lawyer therefore begins every file by asking the same question the Portuguese authorities asked in 2002: on what written terms, and with what verifiable guarantees, is the requesting state asking for the person, and do those terms satisfy French public policy before any surrender is even conceivable.
B. Why did the Supreme Court refuse early release and remission?
Before the Supreme Court, senior advocate Rishi Malhotra argued that the 25-year term should be calculated from 18 September 2002, when Salem was detained in Portugal following the Interpol notice, and that time spent as an undertrial plus remission earned in prison should count toward the total, so that Salem had already served 26 years, nine months and 22 days. The claim also sought deduction of three years, six months and two days of remission earned in prison from the 25-year period. The Bench rejected each branch of that computation. It held that the assurance could not be read as conferring a right to what it called an artificial or accelerated computation of the period of incarceration, ruling that the assurance barred the death penalty and detention beyond 25 years but conferred no right to an artificial or accelerated computation of the incarceration period
The Court gave two linked reasons. First, the starting point is the surrender and the detention that followed in India, not the earlier detention in Portugal before extradition was granted; since Salem was first arrested on 11 November 2005, the 25-year period expires only in November 2030. Second, remission does not shorten the cap, because the sentence itself remains life imprisonment and the 25-year limit flows solely from the sovereign assurance. The Court stressed that Salem stands convicted to life imprisonment as imposed by the competent courts and affirmed on appeal, and the 25-year limit is a consequence of the sovereign assurance given by the Union government to Portugal which does not convert life imprisonment into a fixed 25-year term
That reasoning has a sharp practical edge. The state government’s computation chart placed his probable date of release at 31 January 2046 on a consecutive-sentence reading, while the courts’ headline computation from the November 2005 arrest points to November 2030 as the outer limit imposed by the assurance, with presidential remission to be advised at that stage under the July 2022 ruling. Either way, the September 2026 decision closes the door on the argument that undertrial detention abroad or prison remission can pull the date forward. Defence teams in cross-border cases should take note: when a surrender depends on a written assurance, every day credited against the cap will be litigated, and courts tend to construe the assurance strictly according to its wording rather than as a general indulgence.
The French parallel on double criminality and thresholds shows the same strict-construction reflex. Article 696-3 of the Code of criminal procedure lists the acts that may give rise to extradition and adds: “En aucun cas l’extradition n’est accordée par le gouvernement français si le fait n’est pas puni par la loi française d’une peine criminelle ou correctionnelle.” Attempt and complicity follow the same rule only where punishable under both laws. A French examining court therefore measures the foreign file against French criminal law before it ever reaches assurances, sentence ceilings or remission, and a mismatch at this stage ends the case. For a person sought by a foreign state, the file to build from day one includes the foreign arrest warrant, the exact legal qualification abroad, the French equivalent offence, limitation calculations under both laws, and any assurance letters, because each layer is tested separately and a failure on one is enough to block surrender.
II. Would France surrender a fugitive on the same kind of promise?
France would recognise the mechanism instantly, because French courts deal with conditional surrender all the time, but Paris applies its own filters and they are demanding. The question is never only whether the requesting state promises something. It is whether the promise is specific to the person, verifiable, compatible with French public policy and the European Convention on Human Rights, and embedded in a procedure where the person sought is heard, assisted by counsel, and entitled to challenge each link. The Abu Salem file is a useful stress test for that framework: a death-penalty assurance, a sentence-cap assurance, a specialty question, and a dispute about how time is counted.
A. Can France extradite when the death penalty is at stake?
The starting point in Paris is refusal. Article 696-4 of the Code of criminal procedure lists the cases where extradition is not granted, including French nationality assessed at the time of the offence, political offences or requests made for a political purpose, offences committed on French territory, cases already finally judged in France, limitation, penalties contrary to French public policy, trial by a court not ensuring fundamental procedural guarantees and defence rights, and purely military offences. The death-penalty filter sits at the intersection of several of these heads, reinforced by the Constitution, Protocol No. 13 to the Convention, and Article 3 case law: France does not surrender a person to face a real risk of execution without an assurance that the penalty will not be imposed or, if imposed, will not be carried out, and even then the assurance must be concrete.
The Cour de cassation polices that concreteness strictly. In a 30 January 2024 ruling on a Tunisian extradition request, case number 23-83.549, the criminal chamber restated the governing principle: “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 full holding was issued under Article 3 of the European Convention on Human Rights and Article 696-15 of the Code of criminal procedure, with the Court adding that the investigating chamber’s decision must meet the essential formal conditions of its legal existence and that inadequate or contradictory reasoning counts as no reasoning. On the facts, the chamber of Aix-en-Provence had relied on general considerations about a thirty-year moratorium in Tunisia and renewed international commitments described as symbolic, and the Court quashed the favourable opinion because those generalities did not establish that the death penalty, if pronounced, would not be applied specifically to the person concerned.
The lesson for any requesting state is direct: a note verbale with generalities about a moratorium or treaty membership does not suffice in Paris. The chamber of instruction had ordered additional information precisely so that, if a death sentence were pronounced, the Tunisian authorities would undertake not to have it carried out, and the reply received on 21 March still failed the test of individual effectiveness. Transposed to an Abu Salem-type file, a French court would ask whether the cap or non-execution promise names the person, binds the competent authority, covers every procedural stage including retrial or resentencing, and can be verified after surrender. A promise addressed to the world in general, or conditioned on a future discretionary pardon, would face the same fate as the Tunisian note: cassation and remand to a differently composed chamber.
Procedure matters as much as substance. Article 696-26 of the Code of criminal procedure requires the prosecutor general, within two days of arrest, to notify the person in a language she understands of the documents underlying the arrest, to advise her that she may consent to extradition under the simplified procedure and that she may waive specialty, with mention in the record on pain of nullity, and with the right to counsel. Article 696-15 of the Code of criminal procedure then provides that where the person tells the prosecutor general she does not consent, the investigating chamber is seized without delay, the person appears within ten working days, and the chamber gives a reasoned opinion within one month of appearance unless further information is ordered, with an unfavourable opinion where legal conditions are unmet or there is manifest error, and with cassation available only on formal defects depriving the opinion of the essential conditions of its legal existence. In practice this means the defence has weeks, not months, to file evidence on the death-penalty risk, prison conditions, judicial guarantees, limitation, nationality and specialty, and every exhibit should be translated, dated and sourced, because the chamber rules on the file as completed, and the Cour de cassation will not reweigh facts but will quash for defective reasoning.
B. What happens after surrender: how far does the specialty rule protect?
Extradition in France, as in the Salem file, never authorises a free-floating prosecution for everything the person may once have done. The specialty principle confines the requesting state to the offences for which surrender was granted, for acts predating surrender. Article 696-6 of the Code of criminal procedure states the rule plainly: “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.” The person may waive specialty, but Article 696-26 of the Code of criminal procedure requires that she be informed of that option with counsel and that mention appear in the record, so a waiver is never presumed from silence or from consent to surrender alone.
The Cour de cassation draws concrete consequences from this text even inside France when the person was surrendered to France. In a 12 April 2022 decision, case number 22-80.284, published in the Bulletin, the criminal chamber held under Articles 696-6 and 593: “lorsque la juridiction qui a émis le mandat d’arrêt a obtenu la remise de la personne recherchée, celle-ci ne peut être poursuivie, condamnée ou détenue en vue de l’exécution d’une peine privative de liberté pour un fait quelconque antérieur à la remise et autre que celui qui a motivé cette mesure.” Where the person claims the surrender decision is missing from the file, the investigating chamber must order it to be added and then verify whether pretrial detention was ordered for counts covered by the surrender, in whole or in part. For defence counsel this is an operational checklist: obtain the foreign surrender decision, compare each count with the detention order or indictment, contest any detention resting on a non-surrendered pre-surrender fact, and raise specialty before the investigating chamber rather than waiting for trial, because the chamber is the judge of that compliance.
Inside the European Union the same idea operates through the European Arrest Warrant, which is faster but not looser. Article 695-11 of the Code of criminal procedure defines the warrant as a judicial decision issued by one Member State for arrest and surrender by another of a person sought for prosecution or for execution of a custodial sentence or security measure. Refusal grounds are codified, including Article 695-22 of the Code of criminal procedure, which requires refusal where the acts could be prosecuted in France but are covered by amnesty, where a final decision for the same acts already exists in France or another Member State with the sentence executed or unenforceable, where the person was under thirteen at the time, or where the warrant was issued for discriminatory purposes linked to sex, race, religion, ethnic origin, nationality, language, political opinions, sexual orientation or gender identity. Specialty under Article 27 of Framework Decision 2002/584 and Article 695-18 follows the same structure as in classic extradition, with consent and notification mechanics adapted to mutual recognition. The Lyons saga of 2026, in which a Scottish suspect arrested in Bali was routed through the Netherlands before surrender to Spain on a Spanish warrant, illustrates the plumbing: where no direct surrender treaty exists with the state of arrest, the person transits through the executing Member State whose court tests the warrant, and the issuing state must then respect the scope of what was granted, including any murder count added later only through the proper consent procedure.
Three practical consequences follow for anyone facing a cross-border file with French connections. First, identify the exact surrender title: classic extradition outside the Union or EAW inside it, because thresholds, time limits, refusal grounds and specialty mechanics differ. Second, treat every assurance as evidence to be tested: who signed it, does it name you, does it bind the sentencing and executing authorities, does it cover retrial, resentencing and consecutive sentences, and what post-surrender monitoring exists if it is breached. Third, count time from the correct start date under the correct law: Salem lost because he counted from Portugal detention and deducted remission, while the Court counted from the Indian arrest and treated remission as irrelevant to a cap that is not a sentence. In a French file the equivalent traps are limitation under two laws, credit for foreign detention, and the effect of consent or waiver on detention, and each must be documented with certified translations, because the investigating chamber decides on papers within a one-month horizon and the Cour de cassation reviews only formal legality.
Conclusion
The Abu Salem ruling of 10 September 2026 is a textbook on the long life of extradition promises. Portugal’s 2005 surrender, secured by a December 2002 assurance of no death penalty and no more than 25 years, still governs Salem’s detention two decades later, and the Supreme Court has now fixed the outer limit at November 2030 while refusing to accelerate it through backdating or remission. The life sentences for the Pradeep Jain murder and the 1993 Mumbai blasts stand, the cap operates only as a boundary, and presidential remission will fall to be advised when the 25 years are complete. France would approach an equivalent file with the same seriousness but through its own gates: double criminality and thresholds, mandatory refusal grounds, a concrete and person-specific death-penalty assurance tested against Article 3, a reasoned investigating-chamber opinion within one month, cassation confined to formal defects, and a specialty rule that bars prosecution or detention for any other pre-surrender act without consent. Anyone sought across borders should therefore build the file early, in both languages, around the warrant, the assurance letter, the surrender decision, limitation charts and detention records, because once surrender happens the promises travel with the person, and courts on both sides will hold the parties to the exact words that were written.