The first claimant, Shaun Thompson, was wrongly matched by LFR to his brother, who was on the police watchlist, and as a result of that mistaken identification, Mr. Thompson was questioned and threatened with arrest [17, 141-142]. He was joined in the challenge by Silkie Carlo, director of the civil liberties group Big Brother Watch. The claimants’ case was not that LFR is inherently unlawful. It was that the Policy gives police officers excessive discretion over where, why, and against whom the technology is deployed, thereby breaching Articles 8, 10 and 11 of the European Convention of Human Rights (ECHR). The Policy confines deployment to three primary use cases: (i) crime and missing-person hotspots, (ii) protective security operations, and (iii) specific-intelligence operations. For each, it further sets out the categories of people who may be placed on the watchlist, ranging from those suspected of offences relevant to the hotspot to those wanted by the courts or subject to court orders [110, 113, 115]. The Equality and Human Rights Commission (EHRC) intervened, submitting that the watchlist composition (who) was largely disconnected from the location (where) and objective (why) of deployment, producing “indiscriminate and arbitrary processing”. The Court did not agree. The why, who, and where questions, it held, “form interlocking parts of a Policy that does not authorise the making by the police of arbitrary decisions” [217]; read “fairly and as a whole” [227], the Policy was “a far cry from the “hunch” or “professional intuition” of an individual officer” [200].

The Question the Court Chose

The framing of the issues was the defining feature of the case. The Court held that a challenge based on foreseeability “does not involve a challenge to the proportionality of the measure” [14]. The claimants rather pleaded the former: whether the Policy had the required “quality of law”, and did not argue necessity or proportionality. That meant the Court’s decision was limited. It did not decide whether scanning large numbers of faces was itself justified. Foreseeability asks whether people can know the rules under which they are surveilled; necessity asks whether that surveillance should happen at all. By resolving the case on the first question alone, the Court reviewed the rulebook, but left the practice untested.

The Missed Opportunity

This is where the Case, despite its careful reasoning, falls short. The way a question is framed can shape the answer it receives, and here the framing foreclosed the deeper inquiry.

First, the Court was content to treat a police officer’s self-authored operational policy as the law that authorises indiscriminate biometric capture. The Court upheld LFR without any dedicated facial-recognition statute, and instead accepted a patchwork framework [45, 64], for the operational detail of who, where, and why. The foreseeability doctrine permits this, since “the rules governing the scope and application of measures need not be statutory, provided they operate within a framework of law” [37]. That conclusion rests on classification inherited from Bridges, that LFR is not an “intrusive” method requiring specific statutory legitimacy [43 – 44]. The result is that a uniquely powerful biometric capability operates without any FR-specific legislation and a parliamentary mandate. This is a democratic-legitimacy gap that the foreseeability test was never built to answer.

Second, the Case does not engage with the Glukhin v Russia, even though the EHRC placed it before the Court. Glukhin had a direct bearing on the “quality of law” limb decided here. It called LFR “highly intrusive” [90], demanded “high level of justification…considered necessary in a democratic society” [86], and warned “the need for safeguards will be all the greater” for LFR [82]. On the same question, as such, Strasbourg went the other way.

Third, foreseeability has a human cost. The judgment records Mr. Thompson’s mistaken identification as a fact, but the Court’s narrow frame leaves that experience with little legal weight. A policy can be clear on paper and still produce damaging errors in practice, especially for communities already exposed to heavier policing. A proportionality analysis would have had to confront that reality. A foreseeability analysis does not.

Conclusion

The case is not wrongly decided on the question it answered. The Policy is clearly more detailed, and more constrained, than the framework rejected in Bridges. Yet that is not the same as saying LFR is compatible with human rights. The judgment approves the legal form of the power while leaving its necessity largely unexamined. The claimants have confirmed they will appeal, and the appellate court may yet ask the question this Court left open: not whether the rules are knowable but whether scanning millions of innocent faces can ever be necessary in a democratic society. Until that question is answered directly, the legality of facial recognition in the UK rests on a narrower foundation than the headlines suggest.

Background and intervention procedure

Under Article 63 of the ICJ Statute, a state that is not party to proceedings has a right to intervene where the construction of a convention to which it is party is in question. Before June 2024, the Rules of the Court did not significantly limit this right. The Court, of course, enjoyed substantive discretion on admissibility, but as a matter of procedure, states could file interventions up until the date fixed for the opening of oral proceedings and still be entitled to oral intervention (Rules of the Court Article 82 and 86(2), pre-reform). Last-minute declarations were procedurally admissible as of right.

Article 82 went even further, allowing declarations after the opening of oral proceedings in “exceptional circumstances”. The right of an intervening state to be heard in oral proceedings was, by any metric, procedurally expansive.

Mass intervention in Ukraine v Russia

This procedure was used to full effect in the “tsunami” of oral interventions in Ukraine v Russia. The thirty-one admissible interventions were unprecedented in number, and the intervenors’ political support for Ukraine was clear: they were overwhelmingly signatories to a Joint Statement welcoming Ukraine’s application, and their interventions may have been prepared collaboratively. Under the old procedure, the Ukraine v Russia interventions were heard at length by the Court in a powerful moment of international solidarity.

However, the interventions also went beyond the construction of the Genocide Convention. Russia objected that the intervenors in fact became de facto co-applicants in the proceedings. While the submission of abuse of process was rejected, Vice President Gevorgian voted in favour of admitting the declarations only on the basis that there was no discretion not to do so where the procedural requirements had been met. He raised concerns that the “scope of what is permissible under Article 63” was being circumvented [1].

Reform of the intervention procedure

In the wake of Ukraine v Russia, the intervention procedure was substantially reformed and Article 63 more tightly circumscribed. In particular:

• Declarations of intervention must now be filed no later than the date of filing of the Counter-Memorial (Rules of the Court Article 82).
• Whether Article 63 interventions will be heard in oral proceedings is now at the discretion of the Court (Rules of the Court Article 86(2)).

In essence, non-parties must not only prepare and file their Article 63 declarations at an earlier stage, but also there is no longer any guarantee that they will be heard in oral proceedings.

Implications and conclusion  

The Court’s new discretion is almost certainly intended to prevent a further incident of political mass intervention. Its full impact will soon be visible in South Africa v Israel, where the trend of numerous interventions has continued. To date, eighteen states maintain their interventions in that case. How many will be heard in oral proceedings, however, remains to be seen. It may be that the Article 63 right is vindicated orally, as the obligation to ensure the correct application of the Genocide Convention is erga omnes partes, as reiterated by the Court in The Gambia v Myanmar.

In light of Ukraine v Russia, however, a somewhat restrictive approach appears more likely. As a matter of procedural rigour, a respondent state should not be compelled to answer a wave of intervenors on issues beyond treaty construction. Further, mass political intervention is not in line with the purpose of Article 63 to promote “unity in the law.” The provision is designed to facilitate uniform jurisprudence as intervening parties are bound by the Court’s interpretation of the relevant treaty.

The growth of non-party intervention could be a welcome signal of trust in international tribunals, particularly where both the number of cases being litigated and states participating at the ICJ is increasing. Where participation rises, however, so too must the Court maintain control of its proceedings.

Posidex Technologies Private Limited built Samagra Vedika for the Telangana state government. The system pulls citizen data from over 30 government databases and creates “360-degree profiles.” Machine learning algorithms then compare records across these databases through a process called “entity resolution” to verify welfare eligibility and flag suspected fraud. Officials must consult the system before approving any welfare application. They retain the formal authority to override algorithmic outputs. In practice, they defer to the machine. An Amnesty International technical investigation published in April 2024 confirmed that the system raises serious human rights concerns. Neither Posidex nor the Telangana government has released any source code or accuracy data.

Article 11 of the ICESCR protects the right to adequate food. India’s National Food Security Act 2013 guarantees subsidised food grains to roughly two-thirds of the population. The Supreme Court of India, in PUCL v Union of India (2001), recognised the right to food as part of the right to life under Article 21 of the Constitution. An opaque algorithm now strips citizens of this entitlement. The state offers no explanation and no notice. Affected individuals receive no meaningful opportunity to contest the decision. Instead, excluded beneficiaries carry the burden of proving the algorithm wrong. This reverses the constitutional duty the state owes its citizens.

Other countries have confronted similar failures. In the Netherlands, the Hague District Court struck down a welfare fraud algorithm in the SyRI case (2020) under Article 8 of the ECHR, holding the system’s opacity violated proportionality. In Australia, courts ruled the Robodebt automated debt recovery scheme unlawful. The government repaid AUD 1.8 billion to over 400,000 people. In Serbia, Amnesty International’s 2023 research showed algorithmic welfare decisions disproportionately excluded Roma communities and persons with disabilities from social assistance. Each case carries the same lesson: opaque algorithms making high-stakes decisions about vulnerable populations demand transparency, oversight, and enforceable remedies.

India has no legal framework for algorithmic accountability in welfare delivery. The Digital Personal Data Protection Act 2023 does not address algorithmic discrimination. The law does not mandate impact assessments. Section 17 exempts government agencies on grounds of national security and public order. Welfare recipients have no statutory route to challenge algorithmic exclusions. Meanwhile, the Samagra Vedika model spreads. Haryana, Maharashtra, Tamil Nadu, and Uttar Pradesh have adopted or are developing similar platforms. In Haryana, a pension algorithm declared thousands of living pensioners dead, terminating their old-age and widow pensions. These individuals, many elderly and illiterate, were forced to prove they were alive.

Three reforms are urgent. First, India needs legislation requiring algorithmic impact assessments before any automated system enters welfare delivery. Second, the government must guarantee affected individuals a right to explanation and a meaningful process to challenge algorithmic decisions, consistent with Article 14 of the ICCPR. Third, private developers of welfare algorithms must face independent audits and transparency obligations. The Council of Europe’s Framework Convention on AI (2024) and the EU AI Act’s classification of welfare-related AI as “high-risk” provide working models. Without these safeguards, India’s expanding digital welfare infrastructure will deny food, pensions, and housing to the people who need them most.

Chief Minister Himanta Biswa has interpreted the Supreme Court (‘SC’) judgment in Re-Section 6A as having allowed application of the Immigrants (Expulsion from Assam) Act 1950 (‘Act’) for facilitating pushback of illegal immigrants. The archaic Act’s use was suspended following the Liaqat-Nehru Pact and subsequently fell into disuse, with the legislative vacuum being filled by subsequent laws. Hence, revival of the Act raises the question of whether it should be brought into force after decades of disuse.

Under Section 2, the Act allows the Central Government to formulate an opinion directing the removal of any person ordinarily residing outside India whose stay is detrimental to the public interest. Section 3 allows the delegation of such function to any officer subordinate to the Central Government. However, the judgment in Re Section 6A, which upheld the validity of section 6A providing for special provisions for citizenship determination for Assam, did not authorise bypassing the judicial procedure of adjudication by Foreigners Tribunals (‘FT’). The majority, writing through Justice Kant, held that both the Immigrants (Expulsion from Assam) Act and the Foreigners Act 1946 “seek to regulate the residence and departure of foreigners in India. To that extent, there is no conflict as both of them supplement and complement each other within the framework of Section 6A” [376] and “the provisions of the Immigrants (Expulsion from Assam) Act, 1950 shall also be read into Section 6A and shall be effectively employed for the purpose of identification of illegal immigrants”. [391] (emphasis added)

However, the state has relied on a stray sentence of the judgment, namely, to read provisions of the Act into section 6A, to provide legitimacy for an arbitrary expulsion regime.  This is merely a fig leaf to hide the effective bypassing of the statutory procedures established in post-1950 legislation. This approach also does violence to the language of the carefully crafted section 6 (A) (1) (b) of the Citizenship Act 1955, which expressly links “detected to be a foreigner” to the Foreigners Act 1946 and the Foreigners’ (Tribunals) Order 1964. The Tribunal Mechanism is rendered redundant through SOP’s alternate mechanism, violating the principle that courts should strive to give meaning to a provision rather than render it void.

The statute, even though historically embedded, is always speaking. The Supreme Court in Zee Telefilms Ltd v Union of India, has referred to  Francis Bennion’s Statutory Interpretation to define ‘updating construction’ as the presumption that parliament intends courts to continuously update an ongoing Act’s wording to reflect changes since its enactment. Applied here, Justice Kant’s use of ‘identification’ should be read as preliminary identification by District Collectors (‘DC’) under section 3 of the Act for further reference to the Foreigners’ Tribunal, which can subsequently verify status. This distinction between referential and adjudicatory authority is clear in State of Assam v Moslem Mandal, wherein the court held that the state, before making reference to the FT, must verify the basis of DC’s suspicion. Mere non-reply to notice served on a suspected foreigner would not lead the tribunal to conclude that that person is a foreigner.

Moreover, in Dhoom Singh v Prakash Chandra Sethi, the SC held that there are two consequences when applying ‘updating construction’. First, the legislature’s intent may not be treated as vain, and second, a statute may not be extended to meet a case for which provisions had undoubtedly not been made. In the present case, the legislature’s intent has been to introduce a judicial element in the eviction process. This is visible from the Assam Government’s “White Paper on Foreigners Issue”, which notes how the Central Government introduced the judicial element following the diplomatic opposition to the original procedure, which effectively led to the deportation of nearly 180,000 alleged infiltrators. As a result, specialised FTs were established reflecting a statutory shift in interpreting the word ‘opinion’ in paragraph 2 (1) of the order, from earlier ‘opinion’ amounting to administrative discretion by Border Security Force, to subsequent opinion rendered by FT, thereby installing a 3-step procedure of Identification, Verification and Deportation. In process, identification involves first, suspicion formation by the Inquiring Authority, and second, prima facie verification by the Referral Authority. Where the suspicion persists, reference is made to the FT for seeking an ‘opinion’. The proposed SOP truncates this intended 3-step procedure of identification, verification, and expulsion to a two-step procedure of identification and expulsion by conflating identification and verification.

In conclusion, there is no denying the fact that there has been illegal immigration to Assam. However, the proposed SOP risks arbitrary expulsion and statelessness, which would not only violate the principle of natural justice, but also is in direct conflict with India’s international law obligations. Further, such a push for administrative efficiency runs counter to the SC’s jurisprudence on maintaining due process safeguards concerning human rights.

This is Part 1 of a two-part blog. Here, I shall examine the arbitrariness of the procedure, leading to erosion of the right to fair hearing, and its conflict with India’s constitutional commitments and international obligations.

Recently, Parichay, NLSIU, and Queen Mary University of London released a joint report on the fragile citizenship determination process in Assam, finding structural deficiencies of the FT, which violated three foundational principles of  (a) Rule of Law, (b) Right to Fair Trial and Equality of Arms, and (c) Principles of Natural Justice, due to their operation under shadow of executive control, which ends up disproportionately harming marginalised communities.

The proposed process, by authorising DC to bypass the FT altogether, extinguishes the hint of judicial oversight previously present in the process of citizenship adjudication. Such a procedure contradicts India’s obligations under customary international law and Article 2 (3) of the International Covenant on Civil and Political Rights (ICCPR), which guarantees the right to an effective remedy and prohibits arbitrary deprivation of nationality.

The foremost issue is that the term ‘suspected foreigner’ is undefined, rendering its scope undetermined. Assam’s National Register of Citizens (‘NRC’) exercise has already fragmented people into sub-categories according to their citizenship status. Consequently, a case could come to the FT in 3 ways: (a) referral by Assam Border Police, (b) the Election Commission’s tagging of ‘doubtful’ voter, and  (c) the appeals process in the NRC (which hasn’t started). Take the example of ‘doubtful voters’: as of 2024, about 3.7 lakh persons were classified as ‘doubtful voters’, rendering their citizenship status in limbo. This denies them the right to vote and legal security and often leads to referral to FTs. Despite the Central Government’s direction to Assam’s Government for issuance of rejection slips highlighting reasons for non-inclusion in NRC list to enable individuals to file appeals, these rejection slips have not seen the light of day. This pre-existing suspicion, coupled with unguided discretionary power given to the DC would have severe consequences. Moreover, the threshold of consideration for DCs is the existence of suspicion over citizenship status of an individual and not the proof of them being a foreigner. This violates the right to dignity of the individuals, forcing them to prove their belongingness to India without clarity on what the law requires them to produce. Not only this, but past experience has shown the denial of corroborative documents and rejection of documents for inaccuracies about facts they were not intended to prove. Such piecemeal reading of evidence prevents the interaction of different pieces of proof, undermining fair adjudication. This contradicts the ruling of the Supreme Court in Md. Rahim Ali @ Abdur Rahim v State of Assam, where it was observed that a ‘mere allegation’ is insufficient to call a person’s nationality into question, as this could have life-altering consequences, and the Foreigner’s Tribunal was mandated to serve the primary material supporting allegations against citizenship.

A second issue is the time duration to complete the process. Significantly, not only a reverse burden of proof is mandated on the individual, but the person also has exactly 10 days to prove his citizenship, without any specification of documents that are required to validate such status. In the context of Assam, this is more complicated given the socio-economic and flood-ridden geographical conditions. Also, upon failure to sustain his case, such person is directed to voluntarily leave the country within 12 hours, which directly violates principles of natural justice. Finally, the entire procedure lacks an appellate mechanism, which leads to direct violation of the right to an effective remedy. The only safeguard is to approach the writ court. However, for an average person of the state, to approach the writ court within 12 hours is improbable. Even otherwise, Guwahati HC’s reliance on certiorari jurisdiction under Article 226 of the Constitution permits intervention only on errors of law, without clarifying what such errors entail, and excludes factual errors with the pretext of avoiding reappreciation of evidence. This self-imposed limitation, justified as deference to the tribunal’s findings, denies individuals the right to an effective remedy against grave factual misjudgements.

Overall, the procedure prescribed is perverse to what Hannah Arendt described as a ‘right to have rights’, by having arbitrary requirements without any framework for judicial oversight and the absence of an appellate mechanism.

In response to years of concern about the vulnerability of unaccompanied children, Congress passed the William Wilberforce Trafficking Victims Protection Reauthorization Act (TVPRA) in 2008. This anti-trafficking law established crucial requirements aimed at reducing the risk of exploitation and abuse. Among its key provisions, the TVPRA distinguishes between children from contiguous and non-contiguous countries. Children from contiguous countries—primarily Mexico—may be repatriated swiftly following a limited screening for trafficking indicators or fear of persecution. In contrast, children from non-contiguous countries must be placed in full removal proceedings, with access to pro bono legal representation to the extent practicable, and allowed to pursue claims for asylum or other forms of protection. The statute also imposes obligations to ensure safe repatriation.

Under the July 2025 law, the U.S. government will now ‘permit’ children from non-contiguous countries to withdraw their applications for admission—in effect, abandon their claims—if a perfunctory screening does not reveal dangers of trafficking or persecution. Consistent with other recent developments, this modification to how unaccompanied children may be processed represents another tool to expedite removals. While not surprising—officials during the Trump administration’s first term identified this as a top priority—it is concerning. The U.S. government reportedly is coordinating with Guatemalan authorities to return upwards of 600 children without first allowing them to present a case, in contravention of the TVPRA. A federal judge, however, has temporarily halted them from moving ahead with the removals. Similar negotiations are expected with other countries.

This change effectively extends to unaccompanied children from non-contiguous countries the same cursory procedures applied to those from Mexico, which have yielded deeply troubling protection outcomes. Immigration advocates have long documented inconsistent and inadequate screening of Mexican children by U.S. border authorities. A U.S. government report echoed these concerns and highlighted high rates of repatriation, even among very young children. At this preliminary stage, children do not have access to an attorney and often receive incomplete or misleading information about their options. Many may feel pressure to sign documents that they do not fully understand, including forms that waive potential claims for protection and in which they agree to return to Mexico.

A series of recent policy shifts have already heightened the risks faced by unaccompanied children. Earlier in 2025, the U.S. government suspended funding for legal services for an estimated 26,000 children—including infants and toddlers. Immigration advocates filed a lawsuit in federal court, which blocked the termination, a decision the government has appealed. In addition, it is significantly more difficult to reunify children with family members and others who can care for them during the pendency of their immigration cases. Revised guidelines require in-person meetings with such individuals, some of whom are undocumented, and permit the presence of law enforcement officials during these sessions. This has created a chilling effect on sponsorship and contributed to quadrupling the average length of time children spend in custody.

Obstacles to ensuring children’s safety and well-being are mounting, and it is imperative to reverse this trend. Unaccompanied children are viewed as a law enforcement problem instead of a vulnerable population with specific protection needs. Returning them to places without first understanding or addressing the reasons they fled is likely to exacerbate existing perils and precipitate additional cycles of displacement. In the most extreme cases, children returned to their countries of origin may be killed, as reportedly occurred recently in Honduras and has been documented elsewhere. It is essential to uphold established procedural protections to guarantee that unaccompanied children have their claims fairly heard and adjudicated.

Six French citizens of African or North African origin, residing in Roubaix, Marseille, Vaulx-en-Velin, Saint-Ouen, and Besançon, denounced “racial profiling” during 2011 and 2012. They therefore sought a conviction for discrimination, violation of privacy, and freedom of movement. They also sought to compel the French authorities to implement measures to prevent discriminatory checks and ensure the traceability of these checks.

These six plaintiffs are part of a group of 13 men who launched a legal battle, denouncing unjustified checks, sometimes associated with frisking, insults, or informal address.

After losing at first instance in October 2013, all the plaintiffs appealed, and in June 2015, the Paris Court of Appeal ruled in favor of five of them, ordering the State to pay €1,500 in damages to each. In November 2016, the Court of Cassation definitively condemned the State in three cases, a historic first. The six men who had not won their case therefore decided to appeal to the ECtHR in 2017.

The decision handed down on Thursday, June 26, 2025 comes at a time when the extent of this racist practice was confirmed by the Defender of Rights, Claire Hédon in a report which states that “Young men perceived as Black, Arab, or North African are four times more likely to have been stopped than the rest of the population, and 12 times more likely to be subject to a more intensive stop”.

This is the first time France has been condemned on this ground. While stating that it was “aware of the difficulties police officers face in deciding, very quickly and without necessarily having clear internal instructions, whether they are facing a threat,” [129] the Court concluded, regarding  Karim Touil, who was stopped nearly 15 years ago, that there exists in his case “a presumption of discriminatory treatment against him, which the government has failed to rebut” [92]. Five other applicants, however, had their applications dismissed. Like the domestic courts, the Court considered that, for the identity checks of five of the six applicants, all of which had had at least one identified legal basis, no individualised prima facie evidence of a differentiation of treatment had been produced by demonstrating the coexistence of sufficiently strong, clear and concordant inferences capable of giving rise to a presumption of discriminatory treatment [122].

Defender of Rights Claire Hédon recommends (p. 43) implementing traceability for identity checks to ensure that those stopped have the opportunity to seek redress, particularly in the event of an allegation of discrimination. She also suggests amending the legal framework for identity checks by specifying in Article 78-2 of the French Code of Criminal Procedure that identity checks must not be based on protected characteristics, and that the reason for the person’s selection must be objectively clarified and stated to the person stopped, wherever possible.

As early as 2021, the Paris Court of Appeal had already condemned the State for gross negligence following the identity checks of three high school students in 2017 by police officers at Gare du Nord. However, it is clear that no action has been taken since then. This state of affairs is all the more worrying in the context where, a week earlier, the Minister of the Interior, Bruno Retailleau, had launched a veritable hunt for exiles in stations and on trains, against which several unions from the world of justice had filed an appeal with the Conseil d’Etat.

The decision is important, but suffers from a real ambiguity: the Court assures, on the one hand, that “identity checks motivated by physical characteristics, associated with a real or supposed origin” are well established in France, and certainly cannot be reduced to “isolated cases”, but disputes that this is a “systemic” or “generalized” practice. The Council of State (where, notably, the French president of the ECHR, Mattias Guyomar, has spent his entire career) had made a comparable analysis in 2023.

This decision represents a crucial step towards justice and accountability. However, it is clear that the Court did not address the structural nature of discriminatory stop and search practices in France. The risk is that this decision could further entrench the practice of arbitrary stop and search, already common in much of Europe, which stigmatizes migrants and other visible minorities, perpetuates stereotypes, and constitutes an ineffective and counterproductive method of policing.

The rule of law as defined by A.V Dicey refers to supremacy of the law as opposed to arbitrary exercise of powers, equality of all men before the law, and protection of individual human rights of people (Cf Lord Bingham, Joseph Raz and Brian Tamanaha). Legal instruments have long been adopted to  provide frameworks for protection of the rule of law such as the 1215 Magna Carta which introduced the right to a fair trial, protection from arbitrary detention, and the requirement that laws must be applied equally, the writ of habeas corpus securing the presence in court of the defendant or criminal suspect, and the Constitution of the United States of America in operation since 1789, a historical turning point in the development of the rule of law where it sought to create a strong central government whilst maintaining an individual’s autonomy. This later led to the American Declaration of Human Rights and impacted the adoption of the 1789 French Declaration of Man and the Citizen. Globally, there is the Universal Declaration of Human Rights, adopted by the United Nations General Assembly in 1948, which enshrines the rule of law in the international law sphere.  However, with the increasing use of AI in decision making, questions must be asked about how it affects the operation of concepts associated with the rule of law, such as transparency, fairness, and explainability, as this technology is no longer a figment of imagination.

Whilst AI plays a key role in development, its operationalisation should be hinged within the constraints of the rule of law and protection of human rights and values. Debates on AI regulation often do not take into account that AI itself has regulatory effects on the legal system which could potentially endanger the adherence to the rule of law in democratic states. The infrequency of discussions between the intersection of technology and rule of law confirm that whilst AI primarily embraces globalisation, the rule of law, for the most part, is still primarily confined to national borders. Thus, there is need for international collaboration on setting legal and regulatory frameworks to improve transparency and fairness of decisions made by AI.  As of 2023, an AI index analysis of legislative records of 127 countries shows that the number of bills containing ‘artificial intelligence’ that were passed into law grew from just 1 in 2016 to 37 in 2022. An analysis of the parliamentary records on AI in 81 countries likewise shows that mentions of AI in global legislative proceedings have increased nearly 6.5 times since 2016.

Recent developments and global initiatives that have been seen as most influential and well-known include Council of Europe’s Framework on AI, which is the first international legal instrument on AI focusing on alignment with the rule of law, and UNESCO’s G7 tool kit on AI and rule of law. Regionally, the EU Artificial Intelligence Act 2024 focuses on risk levels of AI to ensure transparency and accountability. In the US, a proposed 10-year moratorium on state level AI regulations introduced by the House Energy and Commerce Committee in May 2025, has been met with opposition from bipartisan groups and states that have regulated high-risk uses of the technology, arguing that the measure will pre-empt AI laws and regulations passed recently in dozens of states. The (Data Use and Access) Bill which is at the reporting stage in the United Kingdom’s House of Commons has raised concerns on the narrowed scope of Article 22 under the UK General Data Protection Regulation which could allow more significant AI decisions without meaningful human involvement.

By emphasising shared global responsibility, legal and regulatory frameworks can help mitigate the risks of automation bias and preserve the rule of law.  AI must operate within a clear legal framework and remain subject to judicial oversight to ensure that the rule of law is upheld.

The European Court of Human Rights (the Court), in its judgment of 24 April 2025 (Applications nos. 46949/21, 24989/22  and 39759/22), unanimously ordered France to compensate three applicants for failure to fulfil positive obligations to effectively apply a criminal justice system capable of punishing non-consensual sexual acts by minors, taking into account the legal framework then applicable, and its application to the present cases. This was pursuant to Articles 3 (prohibition of torture and inhuman or degrading treatment), Article 8 (respect for private and family life) and, moreover, in one of them (no. 46949/21), Article 14 (prohibition of discrimination) of the European Convention on Human Rights (Convention).

The Court thus partially upheld the three applications of 17 September 2021, 12 May 2022, and 6 August 2022, denouncing French law and practice which had neither provided the applicants with effective protection against rape, nor adequately taken into account their status as minors and their vulnerable situation at the time of the events, nor (for the first and third applicants) promptly fulfilled their obligation to investigate and punish the perpetrators of the reported offences, and had, moreover, exposed the first applicant to secondary victimization and discriminatory treatment.

The ECHR reiterated that even if, in practice, it may prove difficult to prove the absence of consent without “direct” evidence of rape (traces of violence; direct witnesses), the authorities must examine all the facts and rule after assessing all the circumstances, because the investigation and its conclusions must focus primarily on the issue of lack of consent. The successive increases in statutes of limitations and other sliding statutes of limitations require, in any case, an increasingly “indirect” approach to the law of evidence [197]. This challenge is all the greater given that the principles of the presumption of innocence and respect for the adversarial process necessarily remain.

It should be noted that the new definition of rape resulting from the law of April 21, 2024, which gave rise to the new Article 222-23-1 of France’s Criminal Code, which establishes a threshold of non-consent and criminalizes any act of sexual penetration or any oral-genital act occurring between a minor under the age of 15 and an adult, would have justified a conviction in two of the three proceedings submitted to the ECHR.

The ECHR noted that criminal proceedings relating to sexual offenses are often experienced as an ordeal by the victim. It recommended conducting the proceedings in a way that preserves the image, dignity, and privacy of alleged victims of sexual violence, including through the non-disclosure of information and personal data unrelated to the facts.

But can requests for documents, pleadings, and written submissions by the defense be limited based on these criteria?

The ECHR considered it essential that domestic authorities avoid “reproducing sexist stereotypes in court decisions, minimizing gender-based violence, and exposing women to secondary victimization by using guilt-inducing and moralizing language likely to discourage victims’ confidence in the justice system,” especially when the victims are minors [200].

The ECHR emphasized, in light of developments “in the understanding of how rape is experienced, particularly by minor girls,” [202] that they often offer no physical resistance due to various psychological factors or because they fear violence from the perpetrator.

The Court therefore considered that in this case:

• Certain questions relating to the victim’s behavior at the time of the incident can no longer be asked (Application no. 46949/21), and therefore characterizes them as “gender stereotypes that instill guilt and are likely to discredit the applicant’s testimony” [226].
• Certain grounds relating to the characterization of consent, which are based “primarily on her passive behavior and lack of physical opposition without duly taking into account either her particular vulnerability or her psychological state, contrary to current knowledge regarding the behavior of rape victims, particularly when they are young,” are not relevant (Application no. 39759/22) [243].

Additionally, the Court found discrimination based on sex with respect to the concluding grounds of the judgment of the investigating chamber of 12 November 2020, which describe the facts complained of by the applicant in a caricatured and pejorative manner, referring to the “usual success with the female sex [of the firefighters] and [their] sometimes unbridled behavior towards them,” which did not “encourage them to reflect” [31]. The Court noted that these statements were “completely irrelevant,” “ineffective,” “an attack on the applicant’s dignity,” and therefore discriminatory [228].

By noting that the Court of Cassation did not rule on this point despite the invocation before it of Articles 3, 8, and 14 of the Convention and the Istanbul Convention, it may be argued that the ECHR is suggesting a new ground for appeal through a review of the “relevance” or even the “dignity” of the lower court’s reasoning.

This condemnation of France under the heading of secondary victimization should oblige it to protect itself against such attacks and, at the very least, to integrate this notion into its liability for defective functioning of the public justice service based on Article L. 141-1 of the Code of Judicial Organization.

However, the JEP has recently faced criticism for not focusing on the selection of top perpetrators, but adopting maximalist approaches and case-by-case analysis methodologies that tend to focus on the individual involvement of all those who participated in the conflict. This has prevented the authority from adequately responding to the massive and complex nature of the cases, leading to delays that disappoint both former combatants and victims. Moreover, it has increased the risk of impunity, as dedicating resources to individual cases may mean that the JEP cannot timely fulfill its main objective.

To prevent this, it is necessary to explore new perspectives that allow for a more efficient selection of the highest-ranking perpetrators. Mathematics seems to offer some insights.

Indeed, for centuries, various scientists have studied ways to maximize the probability of success when making such decisions. They have sought to identify an algorithm that would allow for the best and most efficient selection in the shortest time possible. It is said that the first to attempt this was Johannes Kepler, who, in 1611, set out to find the best candidate to be his wife. The main obstacle he faced, like the JEP, was time: if he spent too much time thoroughly reviewing all the candidates, he risked losing the best one, as she might not wait for him to finish deciding. The challenge, then, was to determine when to stop reviewing his options. In other words, he needed to find the most efficient way to choose the most likely best candidate without having to examine all the available options.

Since the mid-20th century, the search for the best method to quickly make a selection became popular as a mathematical problem, applicable to any situation, such as choosing a house, a partner, a university major, a financial investment, or a secretary for a company. This latter scenario gave rise to the so-called “Secretary Problem,” a topic of study in optimal stopping theory, that seeks to determine the moment to stop evaluating our alternatives and make a decision. Essentially, it is the same question Kepler tried to solve but applied to a company that wants the highest probability of finding the best secretary among several candidates without interviewing them all. This problem becomes critical as the number of options increases because it takes more time and effort to examine each one, which would make the selection process inefficient.

After several experts attempted to solve the problem, John Gilbert and Frederick Mosteller from Harvard University found the solution: the 37% rule. They discovered that one only needs to follow a simple algorithm to maximize the probability of finding the best option, at least mathematically. The rule is to first examine only 37% of the available options and then choose the next one that is as good or better than those, even if many more remain unexamined.

This model, of course, does not guarantee absolute certainty that the best choice will be made, nor does it consider the myriad variables that exist in the real world. However, it provides an optimal solution for increasing the likelihood of achieving the best outcome while reducing the time and effort required for the selection process. This is particularly important for institutions like the JEP, which have a very limited operational period and cannot afford to prosecute crimes using a case-by-case methodology. Instead, they must optimize their efforts to identify the top perpetrators. Perhaps the 37% rule does not address the complexity of the various scenarios within the armed conflict. However, it underscores the value of knowing when to stop analyzing cases and applying what has been learned to make quicker decisions.