Inside the Grand Chamber · No. 02 · Extradition to the US · Article 3 ECHR
A woman of 63, sentenced to a maximum of 60 years with a minimum of 40. Formally not life without parole. What this teaches about the Sanchez-Sanchez test and about diplomatic assurances.
In 2022 Italy extradited Beverly McCallum to the United States. In Michigan she faced life imprisonment without parole. Italy obtained a diplomatic assurance that this sentence would not be sought, and the Grand Chamber declared her application manifestly ill-founded.1 On 23 May 2024, more than a year and a half after that decision, she was sentenced to a minimum of 40 and a maximum of 60 years’ imprisonment. She was 63 years old.2
The assurance was honoured. There was no life sentence without parole. But with a minimum of forty years, McCallum will become eligible for ordinary parole only at around the age of one hundred, even if her detention since her arrest in February 2020 is taken into account. Exceptional forms of release, such as release on medical grounds, are left aside here. For her, it makes little difference whether the sentence is called ‘life’ or expressed in years.
For the European Court it does make a difference. The test which the Grand Chamber developed in Sanchez-Sanchez v. the United Kingdom for extradition to the United States first asks whether there is a real risk of life imprisonment without parole.3 The Court does not exclude that a sentence expressed in years may be equivalent to it. But the evidentiary threshold is so high that, in the cases discussed here, the real prospect of release was not examined. And anyone who comes close to that threshold can expect a diplomatic assurance excluding life without parole.
This article examines these two developments, which reinforce each other. The first is a blind spot: the test gives little weight to the length of the sentence, even where it covers the rest of a person’s life. The second is a shift in the assessment of assurances: from an examination of what awaits the person concerned to trust in what the requesting State promises. Together, they leave the prohibition in Article 3 absolute on paper, while weakening the review of compliance with it.
The Court could not have known McCallum’s sentence in 2022. That is precisely why it is instructive. What happened after the decision shows what the test and the assurance left open, and what that means for the effectiveness of legal protection in extradition cases.
Since Soering, a Contracting State itself violates Article 3 when it extradites a person where substantial grounds have been shown for believing that he or she faces a real risk of prohibited treatment.4 The prohibition allows no balancing.5
For long sentences, the starting point is the case-law on life imprisonment. In Kafkaris the Grand Chamber held that a life sentence must be reducible de jure and de facto.6 Vinter added that it is incompatible with human dignity to deprive a person of liberty without at least giving him the chance one day to regain it.7 A review must therefore be possible, focused on the prisoner’s progress, and comparative and international law point to a review no later than twenty-five years after the sentence is imposed.8
The Hungarian cases show how seriously the Court takes this time dimension. A review that becomes possible only after forty years renders a life sentence de facto irreducible. The Court so held in 2016 in T.P. and A.T. v. Hungary and repeated it in April 2026 in Éberling and Others v. Hungary.9 A review after thirty years of detention, about twenty-six years after conviction, was accepted in Bodein v. France.10
The chart compares different mechanisms: the review of a life sentence and the expiry of a minimum term before parole. What they have in common is the time that passes before release is first considered. That is the time Vinter looks at.
In Sanchez-Sanchez the Grand Chamber adapted that case-law to extradition. It departed from Trabelsi v. Belgium, in which the Court had applied the Vinter criteria in full to the United States.11 It replaced them with a two-stage test. First, the applicant must show that, if convicted, he or she faces a real risk of ‘life imprisonment without parole’. Only then does the Court examine whether, from the time of sentencing, there is a mechanism for reviewing the sentence.12
The Court justified this adaptation by the uncertainty of a prognosis made before trial, and by the risk that a finding of a violation would make prosecution impossible. It identified the alternatives itself: prosecution in the requested State, or assurances from the requesting State.13 At the same time, it emphasised that the Article 3 threshold in extradition cases is the same as within the State’s own jurisdiction.14
The first stage names one sentence. Yet the phenomenon of the term of years that amounts in practice to life was on the table. In the facts of the judgment, the Court refers to the United States Sentencing Commission: in addition to 153 sentences of life without parole, a further 168 persons received in 2013 a term of years so long that it was in practice a life sentence.15 The letter from the United States Department of Justice in that case itself described such sentences as de facto life.16 In Sanchez-Sanchez itself, that finding played no role in the test.
The Court does not rule out such an equivalence. In Bijan Balahan v. Sweden, the applicant argued that in California he would become eligible for parole only after 61 years and therefore faced a de facto life sentence. The Court examined that argument, but held that he had not shown a real risk of such a minimum, given the discretion of the prosecutor and the judge under the Three Strikes law. Without that law, the minimum would have been seventeen years.17 It recalled that within the member States a review only after forty years violates Article 3. Whether a very long minimum term before parole must be assessed in the same way in extradition cases, it expressly left open, because the risk of such a minimum had not been shown.18 Van Zyl Smit and Seeds read the judgment more broadly: in their view, the Court treated the time before review in extradition cases as a merely procedural matter.19 In his dissenting opinion, Judge Wojtyczek considered that a minimum of 61 years is a life sentence that is de facto irreducible, and that a real risk begins at a relatively low level of probability.20 In Carvajal Barrios v. Spain, too, the Court rejected a real risk. It took into account the uncertainty about the outcome of a trial that had yet to take place, the safeguards of that trial, the possibility of mitigating factors and of an appeal, and the finding that such sentences are highly unusual in federal cases.21
McCallum was prosecuted for first-degree murder, which in Michigan carries a mandatory sentence of life without parole. Her daughter was found guilty of the same facts in December 2021 and was sentenced to that penalty in February 2022.22 The Italian courts authorised the extradition, referring to the possibility of a pardon by the Governor. On 22 April 2021 the Court indicated an interim measure.23
On 3 December 2021 the United States informed Italy by diplomatic note, conveying the undertaking of the Michigan prosecutor, that McCallum would no longer be prosecuted for first-degree murder but for second-degree murder, which carries life with the possibility of parole or a term of years. The conspiracy charge was dropped; the charge of disinterment and mutilation of a dead body was maintained. Italy adopted a new extradition decree, the interim measure was lifted on 21 January 2022, and McCallum was extradited on 8 July 2022.24
The Grand Chamber declared the application manifestly ill-founded. It applied a presumption of good faith to diplomatic notes from a State with a long history of respect for democracy.25 It also relied on the amended extradition decree and on the United States’ obligations under the extradition treaty, and rejected the argument concerning the role of the Governor.26 With the note, the real risk of life without parole disappeared. The question of which sentence would take its place was not addressed. That is understandable: in 2022 that sentence had not yet been determined.
What followed shows what the assurance did and did not do. On 1 April 2024 a jury found McCallum guilty of second-degree murder and of disinterment and mutilation of a dead body. On 23 May 2024 the sentence followed: a minimum of 40 and a maximum of 60 years.27 The Court could not have known that outcome in 2022. It does, however, teach something about what the assurance left open: it excluded life without parole and limited the charges, but it set no limit on the time before possible release.
Compare two prisoners. The first is serving a life sentence in Hungary and can apply for review only after forty years. The Court sees a violation of Article 3 in that.28 The second is extradited to a State where, at the age of 63, she receives a sentence with a minimum of forty years. In extradition cases the question of review arises only once a real risk of life without parole has first been shown. A minimum term expressed in years risks never reaching that threshold, as Bijan Balahan shows.
For both prisoners, the question Vinter asks is the same: is there a realistic prospect of ever being released? For the second, the answer is less favourable than for the first. Yet the protection of the first is much stronger.
In Hayes and Others v. the United Kingdom, the applicants faced, if convicted on all counts, a minimum of life plus twenty-eight years.29 They expressly argued that they faced a real risk of life, or its equivalent, without parole, because several other counts also carried minimum sentences of twenty years which could be imposed consecutively.30 The Court did not answer that argument. It found that the mandatory life sentence on one count sufficed for the first stage, and moved on to the second.31
Together with Bijan Balahan, this shows that the Court does not avoid the question of life expressed in years, but has not yet answered it as a matter of principle.
Four reasons can be distinguished, some stated, some inferred from the structure of the case-law.
The first is the form of the test itself. The case-law on reducibility, from Kafkaris to Vinter, was developed for sentences imposed as life sentences. Sanchez-Sanchez transposed that case-law to extradition and took over the label. In Bijan Balahan, moreover, the Court left open whether a very long minimum term is equivalent to life; according to Van Zyl Smit and Seeds, it treated that period as a procedural matter and thereby kept it outside the test.
The second is restraint towards sentencing in other States. In extradition case-law, long terms of years are assessed against the criterion of a ‘grossly disproportionate’ sentence. That is a strict test, met only on ‘rare and unique occasions’, because sentencing legitimately differs between States.32
The third is the argument against speculation. In his dissenting opinion of April 2026, Judge Pisani wrote that the Court cannot speculate about the time a person is likely to spend in prison, since that depends on how long the person lives.33
The fourth is the consequentialist reasoning of Sanchez-Sanchez: preventing impunity and safe havens.34
None of these reasons holds against the foundation the Court itself gave to its case-law.
Against the form of the test stands the case-law itself. Since Kafkaris the Court has required that a sentence also be reducible de facto, precisely so as not to stop at the label. In Asciutto, in September 2026, it reiterated that a life sentence is compatible with Article 3 only if there is both a prospect of release and a possibility of review, and that the Italian ergastolo ostativo did not meet that standard. The judgment distinguishes the former regime from the period after the 2022 reform, for which no violation was found.35 Anyone who examines the reality of a life sentence cannot ignore the reality of a sentence expressed in years.
Against restraint, the point is not whether forty years is too severe for the offence. That is the question of proportionality, and it may remain strict. The point is a different one: whether the person concerned retains any prospect of release. That is the Vinter question, and it depends not on sentencing but on the time that passes before a first review.
Against the speculation argument, simple arithmetic suffices. A minimum term is not speculation: it is in the sentence or in the statute. Nor is the age of the person concerned uncertain. The Court need not calculate life expectancy to establish that a first release at around one hundred is not a realistic prospect. Moreover, in extradition cases the Court always makes a prognosis, and it expressly accepts that this involves uncertainty.36
Finally, against the consequentialist reasoning stands what the Court itself said: the Article 3 threshold is no lower in extradition cases than at home.37 Recognising life expressed in years does not create a safe haven either. As with life sentences, the requesting State can give an assurance about the minimum term or about a review. That is exactly the way out the Court itself identified in Sanchez-Sanchez.38
As long as the test stops at the label, it invites circumvention. An assurance excluding life without parole allows a sentence that achieves the same result. Consecutive terms of years can do the same. McCallum shows that this is no theoretical risk: the assurance was honoured, and the result is nonetheless a sentence that in all likelihood covers the rest of her life.
The applicant must prove a prognosis about a trial that has not yet taken place, before a judge he or she does not know. In that prognosis, the Court does not take into account the possibility of a guilty plea, but it does take into account the possibility of cooperation with the prosecution and the discretion of the judge.39 In Bijan Balahan, that same discretion worked against the applicant. Judge Wojtyczek saw it, on the contrary, as a threat that made the risk real.40 What reduces the risk is taken into account. What makes it concrete, such as the daughter’s life sentence in McCallum, disappears as soon as there is a note.
In Othman the Court listed eleven factors which, depending on the circumstances of the case, it takes into account in assessing an assurance. They include who gives the assurance and whether that person can bind the State, whether local authorities can be expected to abide by it, whether compliance can be objectively verified, and whether the domestic courts have examined the assurance.41 Little of that analysis returns in McCallum. The Court did take into account the amended extradition decree and the United States’ obligations under the extradition treaty. But the starting point remains a presumption of good faith based on the profile of the requesting State, not on the content of an undertaking given by prosecutors at county level.
This line is not new. In Nivette, as early as 2001, the Court accepted an undertaking by an American prosecutor concerning the charges.42 In King, in 2010, the Court attached no importance to the absence of a diplomatic assurance, because the Australian authorities had provided detailed information on sentencing.43 What is new is the weight given to trust. In McCallum the Court also took into account the amended extradition decree, the treaty obligations and the parole system. But what in Othman was one factor among eleven, the strength of bilateral relations, became the starting point of the assessment.
In 2010 Cox proposed a sliding scale: the greater the risk, the stricter the requirements for the assurance.44 In McCallum the opposite happened. The risk was as concrete as it can be in the same case file, and yet the review of the assurance remained limited.
The note in McCallum excluded life without parole. It limited the prosecution to second-degree murder, which carries life with the possibility of parole or a term of years. It contained no concrete limit on the time before possible release. The result, in 2024, is a sentence of a minimum of 40 and a maximum of 60 years for a woman of 63. That is not a breach of the assurance. It is exactly what the assurance allowed.
Othman requires an assurance to be specific.45 The literature on assurances identifies specificity, transparency and post-transfer monitoring as conditions of reliability.46 An assurance concerning life sentences is specific only if it also states which sentence it leaves open. Anyone assessing an assurance must therefore ask not only what it excludes, but also what it permits.
The note concerns the charges. It does not specify who verifies that the prosecution stays within those limits, nor what happens if it does not. After 8 July 2022 McCallum was outside the jurisdiction of any Contracting State. The Court can still establish the responsibility of the extraditing State and award compensation, as Trabelsi shows. But it cannot impose anything on the requesting State, and it can no longer prevent a breach of the assurance itself.
In Koktysh the Court took into account that Ukraine had not shown how it would monitor compliance with the assurances, given the difficult cooperation with Belarus. Cox inferred from this that the extraditing State should itself have required some form of monitoring.47 The CPT, as Cox reports, calls for a monitoring arrangement with unrestricted visits by independent observers, in private with the person concerned.48 The decision in McCallum does not show that any monitoring was provided for; nor does it mention any examination of such monitoring.
It may be objected that an assurance concerning the sentence, like an assurance against the death penalty, is much easier to verify than an assurance against torture.49 That is true. But verifiable is not the same as verified. Moreover, an undertaking about the charges does not automatically bind the judge who passes sentence or the body that later decides on release.
The assurance came only after the judgments of the Court of Appeal and the Court of Cassation, which had relied on a pardon by the Governor. It emerged while the case was pending before the Grand Chamber, under the pressure of an interim measure. An appeal was lodged against the new extradition decree before the Regional Administrative Court of Lazio. The Court’s decision does not state how that appeal ended.50 Whether a national court examined the new undertaking on the merits before the extradition therefore does not appear. Yet Othman counted such an examination as a factor.51
In its 2004 report, Human Rights Watch criticised the assurances in the then pending case of Shamayev as having been obtained only after the extradition.52 In McCallum the assurance did exist before the extradition, but only after the national courts had ruled on it. It is a different form of lateness, with a comparable risk: it is not certain that the court which knows the file best still assesses the assurance.
Cox pointed out that interim measures do not always prevent States from removing a person.53 McCallum shows another danger: the measure holds the State back until an undertaking is in place, and then disappears. Whether any review follows depends on the extraditing State, and does not appear from the decision.
The solution does not require a new test, only a correct reading of the existing one. The first stage of Sanchez-Sanchez should not ask whether the sentence is called life, but whether it operates as life.
Concretely, a term of years should be treated as equivalent to life without parole where the minimum period before a first possible release, given the age of the person concerned at the expected conviction, leaves no realistic prospect of release. The case-law already provides benchmarks: the twenty-five years of Vinter and Asciutto, and the forty years found to be too long in T.P. and A.T. and Éberling.
Once that threshold is reached, the second stage follows as the Court already applies it: is there, from the time of sentencing, a mechanism capable of assessing the prisoner’s progress?54 In Hayes the Court accepted American compassionate release as such a mechanism, although rehabilitation alone is not, by statute, a ground for release there. It relied on the findings that judges do in practice take rehabilitation into account, that individualised reasoning may be required and that an appeal is available.55 That is more than nothing, but Vinter made rehabilitation the core of the review. All that is settled is that release in order to die does not suffice.56 Hayes, however, concerned the federal system. Whether Michigan, a State with its own parole system, provides a comparable review before a forty-year minimum term expires was not addressed in McCallum and must be examined separately.
Checklist for the extradition file
This article is no reproach to the Grand Chamber. It does contain legal criticism of the test that emerged from Sanchez-Sanchez. In that judgment the Grand Chamber adapted the test for extradition, with a view to international cooperation in criminal matters, without calling into question the absolute nature of the prohibition in Article 3, and it gave express reasons for that adaptation.64 The question is whether that test, in practice, still achieves what Article 3 exists for.
McCallum gives a first answer, not through what the Court decided, but through what happened afterwards. Had McCallum received that sentence in a member State, the Court would have had to answer the Vinter question: at the age of 63, with a minimum of forty years, does a realistic prospect of release remain? At her extradition that question was never asked, and afterwards it could no longer be asked. Three lessons stand out. An assurance that excludes life without parole, but sets no limit on the time before possible release, leaves the real sentence open. The evidentiary threshold of the first stage risks leaving the real prospect of release unexamined before extradition. And after extradition, such an examination can no longer protect the person concerned. Effective legal protection therefore requires an assessment, in advance, of what the sentence may become in reality.
The Court itself has said where that standard lies. As early as Soering it held that the safeguards of the Convention must be interpreted and applied so as to make them practical and effective.65 In Kafkaris it required that a sentence be reducible not only de jure but also de facto. In Vinter it linked that requirement to human dignity: no one may lose their liberty without the chance of one day regaining it. That standard looks at the reality of the sentence, not at its name.
That is where the case-law must return. The real test is not whether a sentence is called ‘life without parole’, but what it actually means for this person: how long before a first review, and does a realistic prospect of release remain? What applies within Europe to a Hungarian prisoner must also apply to a person who is extradited. The same holds for diplomatic assurances: an assurance deserves trust when it removes the risk, not when it merely changes the label.
That return can take two paths. The Grand Chamber can refine Sanchez-Sanchez by reading the first stage functionally, as this article proposes. It can also revisit that case-law, as it did in Sanchez-Sanchez itself when it departed from the approach in Trabelsi, having found that the existing case-law no longer fitted.66 Whichever path is chosen, the aim is the same: to bring the protection of Article 3 back into line with the reality of the sentence.
Links to the Court's case-law open the official text on HUDOC: in English where it exists, otherwise in French.
These notes are general information on Convention law; they do not constitute legal advice and do not create a lawyer-client relationship.
After the final decision of the highest national court, a case may be submitted to us directly, by the client or by their lawyer.