Abstract
Mariupol presents an unusual case in international criminal law: a city-scale atrocity documented with unusual rigor across multiple institutional channels, yet still uncertain in its translation into criminal adjudication. The record is dense: satellite imagery, intercepted communications, witness statements taken in dozens of jurisdictions, geolocated video, forensic studies of medical-infrastructure damage, and mapping of filtration operations. The scale of civilian harm and urban destruction is not seriously disputed. What remains uncertain is whether this record can be turned into verdicts, and whether the institutions doing that work will outlast the political conditions that made them possible. Mariupol’s legal afterlife is therefore a problem of architecture as much as evidence. The evidence is abundant. The system meant to carry it from a ruined city to a courtroom is still being built.
The paper works through that system in five layers: the factual record of the siege; the substantive law that applies; the institutions assembled to do the applying — the ICC, Yale’s Humanitarian Research Lab, the OSCE Moscow Mechanism, the Council of Europe Register of Damage, the Eurojust Joint Investigation Team, and the Special Tribunal for the Crime of Aggression created by the 2025 Council of Europe–Ukraine agreement; the doctrinal arguments that decide what prosecutors can actually prove; and the structural weaknesses that could still break the whole effort. Its original claim is methodological: Yale’s remote-sensing record and Human Rights Watch/Truth Hounds’ testimonial reconstruction should not be read as separate tracks. Together, they show that a testimonial layer is not supplemental to satellite evidence; it is structurally necessary if Mariupol is to become a case that can survive trial. The closing argument is that evidence preservation needs to be treated as strategic infrastructure: funded over a long horizon, made interoperable across institutions, grounded in local knowledge, and planned alongside reconstruction rather than treated as something to address after rebuilding has already begun.
I. Introduction: The Long Arc from Event to Adjudication
Mariupol fell in May 2022 after an eighty-six-day siege. The headline numbers have been repeated often enough: a port city of roughly 430,000, reduced (in the language of the UN Human Rights Office) to a scale of destruction Europe had not seen since 1945. The Drama Theater was hit on March 16. Civilians were sheltering inside; the Russian word for “children” had been painted on the forecourt in letters visible from orbit. Maternity Hospital No. 3 was hit on March 9. By May the city was unrecognizable from above. The dead, on credible estimates, ran into the tens of thousands.1
Less attention has been paid to what happened next. From the first weeks of the siege, a documentation apparatus assembled itself across institutions that had not previously coordinated at this scale. The ICC opened its Ukraine investigation on March 2, 2022, after referrals from thirty-nine States Parties. Yale’s Humanitarian Research Lab was publishing satellite-based damage assessments within weeks. The OSCE Moscow Mechanism was invoked. Eurojust set up a Joint Investigation Team with Ukraine, Lithuania, Latvia, Estonia, Poland, Slovakia, and Romania; the ICC joined as a participant, and the United States later as an observer. National prosecutors in Germany, France, the Netherlands, Sweden, Spain, and elsewhere opened structural investigations under their universal-jurisdiction statutes. By the end of 2024, Mariupol-related evidence was distributed across multiple institutional databases, investigative files, and chain-of-custody systems.2
My argument is that the gap between the volume of evidence and the number of actual prosecutions is not, at bottom, a problem of political will, and it is not a problem of law. The law exists. The evidence exists. What is missing, or unstable, is the procedural machinery in between. Evidence collected in one jurisdiction has to be admissible in another. Witness testimony taken in emergency conditions has to survive a defense lawyer’s cross-examination years later. Satellite imagery has to be authenticated under chain-of-custody procedures that did not exist in their present form ten years ago. The substantive prohibitions are all there: in the Rome Statute, the Geneva Conventions, the Genocide Convention, customary international law. What none of those instruments supplies, by themselves, is the connective tissue that gets lived destruction in a Ukrainian port city into a verdict in The Hague, or Berlin, or Strasbourg.
The argument has to take seriously the doctrinal arguments that decide what prosecutors can actually prove. Command responsibility, indispensable to any leadership-level prosecution coming out of Mariupol, was thrown into doubt in 2018 when an ICC Appeals Chamber acquitted Jean-Pierre Bemba by a 3–2 vote. Head-of-state immunity, equally indispensable, sits awkwardly between the ICC’s Al-Bashir Jordan judgment (no immunity before an international court) and decisions like the French Court of Cassation’s July 2025 ruling in the al-Assad case (immunity holds before national courts). Any Mariupol prosecution that ever materializes will have to work through that terrain.
The paper has eight parts. Part II describes what happened during the siege, working from public institutional sources and verified open-source material. Part III maps the relevant legal frameworks. Part IV looks at the institutions assembled to apply them. Part V takes up the three doctrinal disputes that will shape what gets proved — Bemba, Al-Bashir, Al-Werfalli. Part VI sets out the structural gaps that could break the whole project. Part VII proposes a policy framework. Part VIII concludes.
A methodological caveat. The author is from Mariupol; immediate family members were in the city during the opening phase of the siege. This biographical fact does not displace ordinary standards of analytical rigor, and it has not been treated as a substitute for documentary sourcing. Its function in the paper is narrowly defined: to inform editorial judgments about which features of the institutional record are likely reliable, which absences in the public narrative merit comment, and which institutional descriptions of Mariupol require closer scrutiny. It does not bear on the resolution of legal questions. Personal observation has been confined to a single methodological role, identified explicitly where it appears; the analysis otherwise relies on publicly available sources.54
II. The Factual Record: What Happened in Mariupol
A. The Siege in Outline
The siege of Mariupol began on February 24, 2022, the first day of Russia’s full-scale invasion. Russian forces moving north from occupied Crimea, together with Russian-aligned forces from the so-called Donetsk People’s Republic moving in from the east, converged on the city by the first week of March. Mariupol mattered strategically for three reasons: it controlled access to the Sea of Azov, it bridged the southern front to Crimea, and it contained the Azovstal and Illich metallurgical plants, which would become the last defensive positions of the Ukrainian garrison. The city was encircled by March 2. By the fourth day of the invasion, residents in the western districts, including the historically named Zhovtnevyi District, renamed Tsentralnyi in 2016, had lost electricity, water, gas, and cellular communications. The siege would run until May 20, 2022, when the last defenders at Azovstal surrendered after a final stand that drew sustained international attention.
In the first two weeks, civilian infrastructure was hit systematically, and the hits were recorded in real time. Maternity Hospital No. 3 was bombed on March 9, an attack later corroborated by satellite imagery, by photographs, and by witness testimony. The Drama Theater, sheltering an estimated 1,200 civilians and bearing the word “children” in Russian on its forecourts (the letters large enough to be visible from satellite), was destroyed by aerial bombardment on March 16. Estimates of the dead at the theater alone run from 300 to 600; the Associated Press, in the most thorough open-source investigation conducted, arrived at the higher figure. Apartment blocks, schools, kindergartens, and shelters across the city were hit in patterns that, once mapped against geolocated coordinates, showed no clear correlation with any military objective.12
Amnesty International, in a June 2022 investigation, concluded that the Drama Theater attack was “a clear war crime,” pointing to the visible “children” markings, the absence of any clear military objective in the immediate vicinity, and the apparent use of heavy aerial-delivered munitions against a known civilian shelter. The UN Independent International Commission of Inquiry on Ukraine echoed the war-crimes concern, in more guarded language, in its subsequent reporting.13
B. The Documentation of Medical Infrastructure
The most rigorous publicly available assessment of damage to civilian medical infrastructure in Mariupol was produced by researchers at Yale’s Humanitarian Research Lab together with collaborators at Yale Law School. Their pre-post quasi-experimental study, built on commercially available very-high-resolution satellite imagery cross-referenced with geolocated facility data, identified seventy-three medical facilities in the city and found that fifty-six of them — seventy-seven percent — sustained damage during the siege between February 24 and May 20, 2022. Facility size showed no correlation with damage rates, a finding the authors took to be inconsistent with random or collateral targeting. The implication, framed in the careful language characteristic of the study, was that the damage pattern was “not random” and was “consistent with intentional targeting.”8
The legal stakes of this finding are considerable. Article 18 of the Fourth Geneva Convention provides that civilian hospitals “may in no circumstances be the object of attack.”9
Article 8(2)(b)(ix) of the Rome Statute makes intentional attacks against hospitals a war crime, provided the hospitals are not military objectives. Even where intent cannot be directly proven, customary international humanitarian law requires attacking forces to take all feasible precautions to avoid damaging medical facilities. By documenting a damage pattern inconsistent with random or merely collateral harm, the Yale study strengthens the evidentiary basis for a precautions or targeting argument. A prosecutable case would still require attribution, operational context, and proof of the relevant mental element.10
C. The Filtration System
A second body of evidence concerns the filtration system Russian and Russian-aligned forces set up in occupied Donetsk Oblast. The Yale Humanitarian Research Lab’s August 2022 report identified twenty-one sites used for civilian interrogation, processing, and detention. Filtration activities at these sites included registration, holding, secondary interrogation, and, at some locations, what witnesses described as torture. The report identified objects consistent with mass graves at Volnovakha Correctional Colony No. 120 and at the Olenivka prison facility — Olenivka being the site of the July 2022 explosion that killed roughly fifty-three Ukrainian prisoners of war.4
Legally, filtration is a different animal from the bombardment of the city. Bombardment cases turn on proportionality, distinction, and intent. Filtration cases involve enforced disappearance, torture, and the unlawful transfer of civilians from occupied territory, each a distinct offense under both the Fourth Geneva Convention and the Rome Statute. The procedural infrastructure of filtration — its registration points, holding facilities, interrogation centers, longer-term detention sites — also tends to establish the kind of systematic policy that Article 7 of the Rome Statute requires for individual war crimes to be aggregated into crimes against humanity, which demands proof of a widespread or systematic attack directed at a civilian population.
D. The Deportation of Children
The third body of evidence, and the one that has so far produced the most direct legal consequence at the international level, concerns the forced deportation and transfer of Ukrainian children. Yale Humanitarian Research Lab’s 2023 investigation identified an organized network of at least forty-three Russian re-education and adoption facilities and reported at least 6,000 Ukrainian children held in such sites. Separately, Ukrainian official data has documented more than 20,000 children as deported or forcibly displaced; as of May 31, 2026, the Children of War platform listed 2,198 returned. The exact scale remains contested, but the direction of the evidence is not: the transfers were organized, transregional, and embedded in Russian state and occupation-administration structures.3
The UN Independent International Commission of Inquiry on Ukraine, in its March 2023 report, treated these transfers as a war crime and emphasized the systematic and widespread nature of the program, identifying conduct “consistent with the existence of a system.”14
This documentation contributed to the evidentiary environment behind the ICC arrest warrants issued on March 17, 2023, against Russian President Vladimir Putin and Maria Lvova-Belova, the Russian Commissioner for Children’s Rights. They were the first warrants the ICC had ever issued against a sitting head of state of a permanent member of the United Nations Security Council. The charges were brought under Articles 8(2)(a)(vii) and 8(2)(b)(viii) of the Rome Statute, which criminalize unlawful deportation and transfer of population, with children identified as the protected category. Because the ICC’s warrant applications and underlying evidentiary affidavits are not fully public, the precise weight assigned to any one institutional source cannot be known from the public record. What can be said is narrower and stronger: Yale HRL’s documentation of camp locations and transfer pathways has been publicly described as part of the accountability record on which later international legal action drew.6
One point bears emphasis. Strip away the framing, and the conduct underlying the Putin warrant is exactly the conduct described in Article II(e) of the Genocide Convention, which defines genocide to include “forcibly transferring [of] children of the group to another group.” The ICC has charged it as a war crime rather than as genocide. That choice reflects the higher evidentiary burden of proving the specific intent that genocide requires. But the underlying conduct sits squarely on the doctrinal boundary between the two crimes, and Ukrainian prosecutors have argued publicly that the deportation evidence offers one of the most direct routes to a future genocide allegation.1562
III. The Legal Frameworks: From Conduct to Crime
A. The Rome Statute and Its Limits
The Rome Statute of the International Criminal Court, adopted in 1998 and in force since 2002, criminalizes four categories of conduct: genocide, crimes against humanity, war crimes, and the crime of aggression. The first three are governed by Articles 6, 7, and 8. Aggression, added by the 2010 Kampala Amendments and activated in July 2018, sits in Article 8 bis. For Mariupol, the relevant provisions are mainly in Articles 7 and 8, though aggression hangs behind the entire fact pattern as the meta-crime that authorized the invasion.
The Rome Statute’s application to events in Ukraine rests on a jurisdictional fact that is sometimes misunderstood. Russia is not a State Party: it signed in 2000 and withdrew its signature in 2016. Ukraine, too, was for many years not a full State Party — it signed but did not ratify until October 2024. The Court’s jurisdiction over crimes committed on Ukrainian soil during the period this paper covers rests on two declarations Ukraine lodged under Article 12(3) of the Rome Statute, in April 2014 and September 2015. Those declarations accepted ICC jurisdiction over crimes committed on Ukrainian territory regardless of the nationality of the alleged perpetrator. Ukraine ratified the Rome Statute on October 25, 2024, with entry into force on January 1, 2025, but the Article 12(3) declarations remain the operative jurisdictional basis for the period when the crimes in question were committed.17
Two consequences follow from this jurisdictional architecture. Russian nationals can be prosecuted at the ICC for crimes committed in Ukraine, despite Russia’s non-membership in the regime. The ICC cannot, however, prosecute the crime of aggression in the Ukraine situation, because the Kampala Amendments contain a carve-out in Article 15 bis(5) that excludes aggression jurisdiction over the nationals or territory of non-party States. This is the structural gap the Special Tribunal for the Crime of Aggression against Ukraine, created by the Council of Europe–Ukraine agreement signed at Strasbourg on June 25, 2025, was designed to fill. The Rome Statute’s substantive prohibitions on war crimes and crimes against humanity, by contrast, apply in full to the conduct in Mariupol.18
The Article 15 bis(5) carve-out is worth pausing on. Its text excludes ICC jurisdiction over aggression “committed by [a non-party State’s] nationals or on its territory.” William Schabas has argued that the provision produces what he calls a “two-track” Rome Statute — the gravity of the crime of aggression formally acknowledged, its enforcement constrained to a narrower jurisdictional class than the other core crimes. Ukraine has put that two-track structure under acute pressure. Carrie McDougall, in her authoritative study of the crime, anticipated exactly this pressure and predicted that ad hoc tribunals would emerge to fill the gap where the ICC could not reach. The Special Tribunal for Ukraine is that prediction realized.19
B. Command Responsibility
Command responsibility, codified at Article 28 of the Rome Statute and reflected in customary international law, is one of the most analytically important tools international criminal jurisprudence has. Under Article 28(a), a military commander can be held criminally responsible for crimes committed by subordinates under his or her effective control if the commander either knew, or, given the circumstances at the time, should have known, that the subordinates were committing or about to commit such crimes, and failed to take all necessary and reasonable measures within his or her power to prevent or repress them or to refer the matter to the competent authorities.20
The doctrine has a long jurisprudential history. It was applied at Nuremberg and Tokyo. It was refined by the International Criminal Tribunals for the former Yugoslavia and Rwanda — most consequentially by the Tadić jurisdiction decision, which established the customary status of much of international humanitarian law,21
and in the Akayesu judgment, which established both the elements of genocide and the seminal recognition of sexual violence as a constituent act of genocide.22
In the Ukraine situation, command responsibility is the doctrinal hinge on which any leadership-level prosecution will turn. Demonstrating that Russian forces carried out a particular airstrike is one task. Establishing the chain of legal responsibility running from the pilot who released the munition, through the commander who issued the targeting order, up to the political and military leadership in Moscow, is something else. The Bemba case at the ICC, examined at length in Part V, provides both the most extensive interpretation of Article 28 to date and a cautionary illustration of how unstable the doctrine has become.23
C. The Genocide Convention
The 1948 Convention on the Prevention and Punishment of the Crime of Genocide defines genocide as any of a series of enumerated acts committed with intent to destroy, in whole or in part, a national, ethnical, racial or religious group, as such. Article II names five categories of conduct. Killing members of the group. Causing serious bodily or mental harm. The deliberate infliction of conditions of life calculated to bring about physical destruction. Measures intended to prevent births within the group. And, critical for the Mariupol case, the forcible transfer of children of the group to another group.15
The fifth category, forcible transfer of children, offers the most legally direct route to a genocide allegation in the Ukraine context. The same conduct underlies the ICC arrest warrants for Putin and Lvova-Belova, although those warrants charge the conduct as a war crime rather than as genocide. The decision to charge as a war crime is, in part, a function of the evidentiary threshold: the war crime requires proof of unlawful transfer, while genocide additionally requires proof of specific intent (dolus specialis) to destroy a protected group. Several states, among them Ukraine, Poland, Estonia, Latvia, Canada, Lithuania, the Czech Republic, and Ireland, have officially recognized Russia’s conduct in Ukraine as genocide, and the Parliamentary Assembly of the Council of Europe has identified the forcible transfer and so-called russification of Ukrainian children as evidence of genocide.16
The Krstić Appeals Chamber’s reading of the “in whole or in part” element of the genocide definition bears directly on Mariupol. In Krstić, the ICTY upheld a finding that the destruction of the Bosnian Muslim population of Srebrenica — a “substantial part” of the protected group within a specific geographic area — could support a genocide conviction. Applied to Mariupol, the Krstić framework would not require proof of an intent to destroy the Ukrainian nation in its entirety. It would require proof of an intent to destroy a substantial part of the group, with substantiality assessed by reference to specific geographic and demographic markers. The Office of the Prosecutor of Ukraine has publicly described the deportation evidence as “the most promising way to prove genocide” in the Ukraine context.26
D. The Crime of Aggression
The crime of aggression occupies a peculiar position in the international criminal law architecture. It is the only one of the four core crimes that targets the initial decision to use force, not the conduct of force once initiated. Article 8 bis of the Rome Statute defines aggression as the planning, preparation, initiation, or execution, by a person in a position effectively to exercise control over or to direct the political or military action of a State, of an act of aggression that by its character, gravity, and scale constitutes a manifest violation of the UN Charter. It is a leadership crime in the literal sense: only those at the highest levels of state authority can commit it.28
As noted above, the ICC cannot prosecute the crime of aggression in the Ukraine situation. The leading scholarship on the crime — McDougall’s monograph, Tom Dannenbaum’s work on aggression and the soldier — anticipated exactly the structural problem the Kampala carve-out would create in situations like Ukraine, and both have engaged with the ad hoc tribunal proposal as the most plausible route to leadership-level aggression prosecutions.29
The Special Tribunal for the Crime of Aggression against Ukraine, created by the Council of Europe–Ukraine agreement signed at Strasbourg on June 25, 2025, was designed to fill that jurisdictional gap. Its jurisdiction rests on Ukraine’s own territorial jurisdiction, delegated through the bilateral agreement, with administrative and financial support flowing from an Enlarged Partial Agreement of the Council of Europe adopted by thirty-six states and the European Union at the May 2026 session of the Committee of Ministers in Chişinău. The Tribunal’s Statute takes a flexible approach to in absentia proceedings, permitting investigation and indictment to proceed without the accused, with full trial proceedings to follow upon the accused’s presence — a design choice meant to address the near-certainty that Russian leadership will not surrender to the Tribunal voluntarily.30
The relationship between aggression and the conduct in Mariupol is causal rather than constitutive. The siege was made possible by the prior decision to invade. Aggression, in that sense, is the meta-crime that licenses all the others. A successful prosecution before the Special Tribunal would establish, as a matter of legal record, that the violence which destroyed Mariupol was not a tragic consequence of war but the foreseeable outcome of an unlawful initial decision. That is what distinguishes aggression from the other Rome Statute crimes in legal-political terms: it speaks to the wrongfulness of the war itself, not just to the wrongfulness of particular conduct within it.
IV. The Institutional Architecture: Who Collects, Who Prosecutes
A. The International Criminal Court
The ICC opened its Ukraine investigation on March 2, 2022, after state-party referrals from thirty-nine countries. It has issued six arrest warrants in the Ukraine situation to date: the March 2023 warrants against Putin and Lvova-Belova; March 2024 warrants against Russian military officers Sergey Kobylash and Viktor Sokolov for alleged attacks on civilian infrastructure; and June 2024 warrants against Sergei Shoigu and Valery Gerasimov for alleged international crimes committed between October 2022 and March 2023.31
The ICC’s structural limits are familiar. It has no police force of its own, and depends on the cooperation of States Parties to execute its warrants under Articles 86, 87, and 89 of the Rome Statute. Russia has refused to cooperate and has threatened to prosecute ICC officials. State Party performance on cooperation has been uneven: Mongolia’s September 2024 reception of President Putin, despite the warrant, illustrated the practical limits of enforcement, and the ICC found Mongolia in non-cooperation — without imposing meaningful consequences.32
The Mongolia episode was a compressed reprise of the pattern that has frustrated ICC enforcement since the Al-Bashir warrants of 2009 and 2010. States that are formally bound to cooperate find ways to avoid arrest. The institutional sanctions available to the Court (referral to the Assembly of States Parties or the UN Security Council) operate at a political register that usually produces no enforcement consequence.33
B. The Yale Humanitarian Research Lab and the Conflict Observatory
The Yale School of Public Health’s Humanitarian Research Lab, led by Nathaniel A. Raymond and Kaveh Khoshnood, has become one of the most consequential institutional actors in the Ukraine evidence architecture. The Lab operates under the Conflict Observatory, a program funded by the U.S. Department of State’s Bureau of Conflict and Stabilization Operations and partnered with Esri, the Smithsonian Cultural Rescue Initiative, Planetscape Ai, and others. Its methodological signature is the combination of very-high-resolution commercial satellite imagery with open-source intelligence drawn from social media, regional reporting, and public databases — all maintained under formal chain-of-custody procedures designed for use in accountability mechanisms.34
The Lab’s output has shaped the Ukraine accountability landscape on three fronts. It has supplied real-time documentation that has fed into ongoing investigative work at the ICC, the OSCE, and national prosecution offices. Its public reports have set a methodological baseline for what rigorous open-source documentation looks like, and that baseline has been adopted by other institutions. And through the Putin and Lvova-Belova warrants, the Lab has shown that academic open-source documentation can be admitted as the evidentiary foundation for international criminal proceedings — a point Raymond has himself emphasized in public testimony and in his academic writing.35
The Lab’s position is also fragile. Its funding, like much of the U.S. Conflict Observatory program, has been the subject of political contestation. Public reporting in 2025 indicated that significant portions of congressionally approved funding for the Lab’s work had been withheld by the executive branch, and the Lab’s leadership warned publicly of potential closure. That fragility is itself a structural feature of the evidence architecture: much of the documentation on which future prosecutions will rest is being produced by institutions whose continued existence is not guaranteed.36
C. The OSCE Moscow Mechanism
The OSCE’s Moscow Mechanism, invoked in March 2022 by forty-five participating States in connection with Russia’s invasion of Ukraine, produced a comprehensive report on violations of international humanitarian and human rights law. The mechanism is unusual in international institutional design: a group of OSCE participating States can invoke a fact-finding mission without the consent of the state under investigation. The resulting April 2022 report was one the Yale Humanitarian Research Lab supplemented with its own detailed analysis of five Ukrainian cities and regions, including documentation of widespread and systematic bombardment of healthcare facilities.11
The Moscow Mechanism report is not itself a prosecutorial instrument. But it is an authoritative international fact-finding record that can be cited in subsequent proceedings, at the ICC and in national courts. Its evidentiary significance is structural: it locks in a baseline factual record at an early stage of the conflict, against which subsequent disputes about the scope and character of violations can be measured.37
D. The Register of Damage and the International Claims Commission
In May 2023, the Council of Europe set up the Register of Damage Caused by the Aggression of the Russian Federation against Ukraine, an institutional mechanism for recording claims of damage caused by Russia’s internationally wrongful acts. The Register’s scope is broad: damage to individuals, businesses, and the Ukrainian state. By April 2026, Council of Europe reporting indicated that almost 150,000 claims had been submitted and more than 45,000 had been recorded.38
A subsequent treaty, the Convention establishing an International Claims Commission for Ukraine, was concluded in 2025 to provide the adjudicative mechanism for the claims recorded by the Register. The Register and the Commission are not criminal mechanisms; they operate in the civil-reparations space rather than the criminal-prosecution one. They are included here because they illustrate a key feature of the Ukraine evidence architecture: it is multi-modal. Criminal prosecution, civil reparation, and historical documentation each have their own institutional mechanisms, and each draws on overlapping but distinct evidentiary records. The strength of the overall architecture is precisely this redundancy.39
E. National Prosecutions Under Universal Jurisdiction
A growing number of European states have opened structural or specific investigations into Russian conduct in Ukraine under universal-jurisdiction statutes. Germany, France, Lithuania, Latvia, Estonia, Poland, Slovakia, Spain, Sweden, and the Netherlands are among them. The legal theory underlying these prosecutions varies. Some rest on universal jurisdiction in the strict sense, the view that certain crimes are grave enough that any state may prosecute them regardless of territorial or nationality nexus. Others rest on narrower statutory bases — the presence of the alleged perpetrator on the prosecuting state’s territory, or the presence of victims among its residents.55
The German Federal Public Prosecutor’s Office has been particularly active here, drawing on the experience built up through its prosecutions of Syrian war crimes. The 2022 judgment of the Higher Regional Court of Koblenz in the Al-Khatib case — the first criminal verdict anywhere in the world against a former Syrian state official for crimes against humanity — established procedural precedents that have since been applied to Ukraine-related investigations. The German model emphasizes “structural investigations,” which build up a body of evidence about the systematic conduct of a state or armed group as a foundation for subsequent individual prosecutions. The model suits the Mariupol case well: the pattern of strikes on protected infrastructure is the central evidentiary fact.41
The Koblenz model is not without difficulty. The Al-Khatib trial took several years, depended on extensive cooperation between German federal prosecutors and German-based Syrian witnesses, and required the development of a novel evidentiary infrastructure linking structural findings to case-specific ones. Whether the model scales to the volume of evidence generated by the Ukraine conflict — which already dwarfs the Syrian-origin record in raw volume — is an open question. Christian Ritscher, who served as Head of the German War Crimes Unit (S4) during the Al-Khatib period and later as UN Special Adviser and Head of UNITAD, has written about the procedural challenges of running international crimes trials at scale.40
F. The Joint Investigation Team and the ICPA
The Joint Investigation Team on alleged core international crimes committed in Ukraine, set up by Eurojust on March 25, 2022, brings together prosecutors and investigators from Ukraine, Lithuania, Poland, Estonia, Latvia, Slovakia, and Romania, with the ICC as a participant and the United States as an observer. In 2023, the Team was expanded to include the International Centre for the Prosecution of the Crime of Aggression against Ukraine, hosted at Eurojust in The Hague. In European Union law, joint investigation teams are a flexible mechanism for cross-border coordination of criminal investigations: prosecutors and investigators from multiple jurisdictions can share evidence in real time under shared procedural protocols. For a case as large and as evidentially complex as Ukraine, the JIT is the institutional substrate that lets national prosecutions, ICC proceedings, and the eventual work of the Special Tribunal draw on a coordinated rather than fragmented evidentiary record.42
V. The Doctrinal Battlefield: What the Cases Will Turn On
If Part IV mapped the institutional plumbing of the Mariupol evidence architecture, Part V turns to the doctrinal questions those institutions will have to litigate. The substantive law surveyed in Part III applies to Mariupol; the institutions surveyed in Part IV have the capacity to apply it. What remains is the harder question of what, at trial, prosecutors will actually be able to prove. Three case-based controversies dominate the field: Bemba on command responsibility, Al-Bashir on head-of-state immunity, Al-Werfalli on the admissibility of digital open-source evidence. Each functions as a precedent in one direction and as an obstacle in another.
A. The Bemba Problem: Command Responsibility Destabilized
Prosecutor v. Bemba Gombo is the ICC’s most extensive engagement with command responsibility, and the most cautionary. The case concerned the criminal responsibility of Jean-Pierre Bemba, then vice-president of the Democratic Republic of Congo and president and commander-in-chief of the Mouvement de libération du Congo (MLC), for war crimes and crimes against humanity committed by MLC forces during their intervention in the Central African Republic between October 2002 and March 2003. In a 2016 judgment, the Trial Chamber convicted Bemba under Article 28(a) as a remote commander who exercised effective control over the troops and failed to take all necessary and reasonable measures to prevent or repress the crimes.23
On June 8, 2018, a sharply divided Appeals Chamber, by a 3–2 majority, reversed the conviction and acquitted Bemba. The majority’s reasoning matters for any future leadership prosecution. The Trial Chamber, the Appeals Chamber held, had erred on two principal grounds. It had relied improperly on criminal acts that fell outside the scope of the charges as confirmed by the Pre-Trial Chamber. And, more substantively, it had failed to give adequate weight to Bemba’s status as a “remote commander” operating outside his own state, and had not properly assessed what additional measures he could realistically have taken from that position to prevent or repress the crimes. The dissent, by Judges Sanji Mmasenono Monageng and Piotr Hofmański, ran to 269 pages, substantially longer than the 80-page majority judgment, and argued that the majority’s narrowing of Article 28 amounted to “judicial law-making.”24
The critical scholarly response was sharp. Leila Sadat called the judgment a “destabilizing aberration.” Diane Marie Amann wrote that the Appeals Chamber had effectively reduced Article 28 to an “evidentiary cul-de-sac.” Susana SáCouto and Patricia Sellers focused on the disproportionate impact on prosecutions of sexual and gender-based crimes, where command responsibility has historically been the most effective doctrinal route.25
The Bemba problem for Mariupol is acute, but the scholarly response, I think, has overstated it. Sadat’s “destabilizing aberration” framing assumes the Appeals Chamber narrowed Article 28 as a doctrinal matter. Reading the majority opinion closely suggests something else: what the Chamber narrowed was not the elements of command responsibility but the evidentiary requirements for proving them. The Trial Chamber, on the Appeals Chamber’s account, failed to show with the necessary specificity that Bemba had taken inadequate measures *given what he actually knew* as a remote commander. That is a high bar. It is not a categorical bar to remote-commander liability. The distinction matters for Mariupol because the documentary record on Russian command structures looks nothing like the record on Bemba’s MLC. Russian targeting orders flow through a more centralized, better-documented chain than the MLC chain ever did. Intercepted communications, leaked targeting protocols, and the satellite-confirmed pattern of repeated strikes on the same protected categories — hospitals, theaters, shelters — supply exactly the kind of specificity the Bemba majority demanded. Any leadership-level prosecution from the siege will still have to demonstrate effective control, knowledge or constructive knowledge, and failure to act. After Bemba, each element is contested. But the documentary record the Yale Humanitarian Research Lab has assembled, taken together with intercepts and Ukrainian intelligence material now starting to enter the public domain, plausibly clears the post-Bemba bar. The defense will deploy a remote-commander argument; the answer is that the systematicity of the strike pattern is itself evidence of knowledge and acquiescence.
The defense will push back on this argument, and the push-back deserves a direct answer. Counsel for a senior Russian commander would not contest the existence of the strike pattern; they would contest its evidentiary value for the knowledge element. The pattern shows what happened, the argument runs, not what the accused knew was happening, and the inferential leap from systematicity to individual knowledge is exactly the leap Bemba forbids. There is real force here. The Bemba majority did flag the danger of inferring individual mental states from aggregate patterns, particularly where the chain of command stretches across geography. But the argument only cuts so far. The Mariupol record is not aggregate in the way the Central African Republic record was; it is anchored in specific, time-stamped, communications-level material — intercepts in which named commanders discuss specific targets, named units, named outcomes. That material does not require the inferential leap from pattern to knowledge. It supplies knowledge directly. Where it exists, Bemba poses no obstacle. Where it does not, the defense argument retains its force, and the prosecution will have to fall back on the aggregate-corroboration logic discussed in Part V.C below. What Bemba creates, in practice, is a sorting problem: which strikes are documented to the communications level, which are documented only to the pattern level, and which command-responsibility theories survive each sorting. That sorting is not yet public for the Mariupol case. It will be the first analytical task any prosecution team faces.
The point gets sharper when you look at recent work on the Russian order of battle in Mariupol. The Human Rights Watch and Truth Hounds 2024 report identifies, through review of unit obituaries, official decorations, Russian state media, and verified social-media imagery, seventeen distinct Russian and Russia-affiliated units operating in the city during the assault, and names ten individuals up successive levels of the chain of command for whom command-responsibility analysis is at least prima facie viable. The methodology is open and replicable: every unit identification traces to multiple corroborating sources, and the report is candid about the limits of its evidentiary base. What matters for the post-Bemba doctrinal question is not the list itself, but the fact that such a list can be assembled at all, on open sources, for a campaign of this scale. The Bemba majority complained, in effect, that the Trial Chamber treated remote-commander knowledge as an inference from circumstance rather than a fact in the record. The Mariupol record, as it stands in the open sources alone, supplies the kind of unit-level and command-level granularity the Bemba majority indicated it wanted to see.70
B. The Al-Bashir Problem: Head-of-State Immunity in Contested Doctrine
Prosecutor v. Al-Bashir, the long-running ICC proceedings concerning former Sudanese President Omar Hassan Ahmad Al-Bashir, produced the most consequential statement of international criminal law on head-of-state immunity since the Rome Statute entered into force. On May 6, 2019, in the Jordan Referral Re Al-Bashir Appeal, the ICC Appeals Chamber held that “[t]here is neither State practice nor opinio juris that would support the existence of Head of State immunity under customary international law vis-à-vis an international court.” The Chamber further held that “the absence of a rule of customary international law recognising Head of State immunity vis-à-vis international courts is relevant […] also for the horizontal relationship between States when a State is requested by an international court to arrest and surrender the Head of State of another State.”49
The Al-Bashir judgment is doctrinally aggressive. It positions the ICC as an “international court” before which no head of state — of any state, party or non-party to the Rome Statute — enjoys personal immunity. The implications for the Ukraine situation are immediate. President Putin, like President Al-Bashir before him, is the sitting head of state of a non-party. The ICC has issued an arrest warrant against him. Read straight, Al-Bashir removes personal immunity as a defense to that warrant before any of the 125 States Parties to the Rome Statute. That is the doctrinal foundation the Mongolia non-cooperation finding rested on.
The Al-Bashir position is not universally accepted, however. The French Court of Cassation, sitting in plenary assembly on July 25, 2025, ruled on the validity of an arrest warrant issued in November 2023 by French investigating judges against then-Syrian President Bashar al-Assad for complicity in crimes against humanity and war crimes. The Court reaffirmed that the personal immunity (immunity ratione personae) of sitting foreign heads of state remains absolute under customary international law, and that international crimes, crimes against humanity and war crimes included, do not by themselves create an exception to that immunity in proceedings before national courts. The ruling is narrower than a general shield for former officials: because Assad was no longer in office by the time of the judgment, the Court also indicated that new warrants could be issued on functional-immunity terms. The Statute of the Special Tribunal for the Crime of Aggression against Ukraine takes a position closer to the ICC’s Al-Bashir reasoning, providing that personal immunity does not apply in its proceedings; that position is itself the subject of ongoing legal dispute.50
The Special Tribunal’s design choices on immunity and in absentia proceedings deserve sustained attention. Beth Van Schaack and Tom Dannenbaum, U.S. delegates involved in the Core Group meetings that drafted the Tribunal’s statute, have written publicly that the Tribunal’s jurisdiction rests not on the contested theory of an “international court” exception to immunity, but on Ukraine’s sovereign territorial jurisdiction over crimes committed on its territory, delegated through the bilateral agreement to a specially constituted body. In their view, that theory is less doctrinally controversial than the ICC’s Al-Bashir reasoning, and provides a stronger foundation for future enforcement. I am not convinced. The delegation theory holds at the level of the founding agreement, but it does not answer the harder question that arises later, when a non-party state — Switzerland, India, Brazil — is asked to arrest a Russian official on a Special Tribunal warrant. The non-party would be asked to override personal immunity it would otherwise recognize, and the only legal basis for doing so is that the Tribunal’s warrant is functionally equivalent to one issued by a true international court. That is exactly the Al-Bashir argument the Van Schaack/Dannenbaum framing was meant to avoid. The French Court of Cassation’s July 2025 ruling makes the point sharper: if a French court will not override sitting head-of-state immunity to honor a national arrest warrant, the question of whether it would do so for a Special Tribunal warrant is, at best, unresolved. The delegation theory is useful as a framing for the Tribunal’s establishment. It does not, on its own, solve enforcement.
If the delegation theory is doctrinally fragile and the Al-Bashir international-court theory remains doctrinally contested, the realistic position requires accepting that legal theory will not, on its own, resolve the enforcement problem. The Tribunal’s legal architecture supplies a valid warrant; whether that warrant is honored by any given non-party state remains, in practice, a political decision constrained but not determined by international law. This is not a weakness of the Tribunal’s design. It is a correct identification of where the enforcement problem actually sits. The doctrinal contest between delegation and international-court theories matters for the eventual ICJ adjudication that may resolve it, and matters for how courts in States Parties reason about cooperation requests. It does not, however, matter for the practical question of whether a Russian official transiting Switzerland or Brazil will be arrested. That question turns on factors that lie outside the legal architecture entirely: domestic political pressure, bilateral relations with Russia, the calculation of the requested state’s own interests. The realistic posture for the Tribunal’s designers is to build the legal foundation on the strongest doctrinal footing available, accept that enforcement will be uneven, and treat each successful enforcement as precedent-building for the next. That is a slower and less satisfying answer than either the delegation theorists or the Al-Bashir maximalists would prefer. It is, I think, the accurate one.51
For the Mariupol evidence architecture, the Al-Bashir problem sets the doctrinal background against which everything else operates. It shapes what is possible in The Hague, what is possible in national courts of States Parties, and what is possible in non-party jurisdictions. Until the ICJ — to which the African Union has previously sought to refer the question for an advisory opinion — speaks definitively, or until customary international law shifts through the accumulation of state practice, every prosecution that touches a sitting Russian official will operate in the shadow of the Al-Bashir controversy.
C. The Al-Werfalli Precedent: Digital Evidence at the Threshold
Prosecutor v. Al-Werfalli, the proceedings against Mahmoud Mustafa Busayf Al-Werfalli, an alleged commander of the Al-Saiqa Brigade in Libya, marked a turning point in the use of digital evidence at the ICC. On August 15, 2017, the Pre-Trial Chamber issued the first ICC arrest warrant based solely on social media evidence. Seven separate videos, posted to Facebook and other platforms, showed Al-Werfalli either committing or directing the summary execution of thirty-three prisoners in and around Benghazi between June 2016 and July 2017. A second warrant followed in July 2018 after additional video evidence of further executions surfaced. Proceedings were terminated in June 2022 following Al-Werfalli’s reported death.44
Al-Werfalli is the foundational ICC precedent for using digital open-source evidence as the basis for an arrest warrant. It has been the subject of extensive scholarly analysis. Lindsay Freeman has documented its significance as a methodological inflection point, arguing that it demonstrated the ICC’s willingness to engage with social-media-derived evidence at the warrant stage. Alexa Koenig, of the Human Rights Center at UC Berkeley, has emphasized the case’s role in driving the development of formal protocols for the verification and authentication of digital evidence — work that culminated in the Berkeley Protocol on Digital Open Source Investigations.45
The relevance to Mariupol is direct. The Ukraine conflict has produced an order-of-magnitude greater volume of digital open-source evidence than the Libyan conflict did, and the institutional infrastructure for processing and authenticating that evidence has matured considerably since 2017. Bellingcat’s geolocation work on the Al-Werfalli videos was, in many ways, the prototype for the open-source investigation methodology that has since been integrated into the work of the Yale Humanitarian Research Lab, the Conflict Observatory, and the ICC’s own investigative practice. The Berkeley Protocol — drafted in collaboration with the UN Office of the High Commissioner for Human Rights and the Human Rights Center at UC Berkeley — sets out the procedural standards by which such evidence is to be collected, preserved, and authenticated for use in international criminal proceedings.
At the same time, the Al-Werfalli case surfaced limitations Mariupol prosecutions will have to overcome. The most consequential is platform-side deletion. When Bellingcat investigators tried to verify the Al-Werfalli videos, they found that much of the original social-media content had disappeared from Facebook and YouTube in the months after the ICC issued the warrant, sometimes through automated content-moderation removal. Rebecca Hamilton has analyzed this problem in detail, arguing that the platforms’ content-moderation regimes are systematically misaligned with the requirements of war crimes documentation, and that the legal infrastructure has not yet caught up.46
A second limitation is the doctrinal gap between the warrant stage and the conviction stage. The Pre-Trial Chamber’s decision to issue the Al-Werfalli warrant rested on the lower “reasonable grounds to believe” standard. The ICC has not yet had occasion to test whether social-media-derived evidence, standing alone, can support a conviction at the higher “beyond reasonable doubt” standard. The Putin and Lvova-Belova warrants, like Al-Werfalli, rest on the lower standard. Whether the underlying evidence can support a conviction at trial, assuming a trial ever takes place, is the unresolved question the Mariupol evidence architecture will eventually have to answer. My own view is that Al-Werfalli has been read too narrowly by its proponents and too broadly by its critics. Proponents treat it as a green light for purely open-source prosecutions. Critics treat it as a one-off, enabled by particularly damning self-incriminating video. Neither framing captures what is doctrinally novel about the decision, which is the Chamber’s implicit acceptance that *aggregate* digital evidence (multiple independent uploads, cross-verified geolocation, a sustained pattern across months) can do the work that single-source evidence cannot, and substitute for the chain-of-custody guarantees that single-source material cannot supply. The Mariupol record, with its thousands of independent contemporaneous uploads from civilians inside the besieged city, fits that aggregate logic almost too well. The conviction-stage question, then, is not really whether digital evidence is admissible, but how a tribunal weighs corroboration density.
VI. Where the System Breaks: The Structural Gaps
A. The Authentication Problem
The first and most consequential structural problem in the Mariupol evidence architecture is the authentication of digital evidence. Open-source material drawn from social media, mobile phones, and the digital exhaust of an information-saturated war does not come with the chain-of-custody guarantees that physical evidence has historically carried. The Berkeley Protocol on Digital Open Source Investigations, published by the UN Office of the High Commissioner for Human Rights in 2022 in collaboration with the Human Rights Center at UC Berkeley School of Law, set the first comprehensive international standard for the collection, preservation, and use of digital open-source information in legal proceedings. The Protocol is now widely cited but unevenly implemented, and even where it is implemented, it does not resolve the deeper epistemological problem the discipline now faces.43
The authentication problem is not theoretical. Defense counsel in any future Mariupol prosecution can be expected to challenge digital evidence on grounds of provenance, integrity, and interpretation. Has the file been altered? Was the metadata preserved? Can the geolocation be independently verified? Were the witnesses who supplied the original material subject to coercion or influence? Generative AI complicates the picture further: convincing fabricated material is increasingly cheap to produce, and the standards by which a tribunal will distinguish authentic from fabricated digital content are still in flux. International criminal tribunals have generally been admissibility-friendly, applying flexible evidentiary standards under the rubric of probative value, but the weight a tribunal assigns to digital evidence remains a function of the rigor with which it was collected.
A deeper version of the problem is that authentication standards built for the pre-generative-AI era are mismatched to the evidentiary environment a Mariupol tribunal will actually face. The Berkeley Protocol was drafted in 2020–2022, before the wide availability of diffusion-model image generation and frontier video synthesis. Its authentication procedures — provenance verification, metadata preservation, multi-source corroboration — remain methodologically sound, but they were calibrated for an environment in which producing convincing fabricated material took substantial technical skill and resources. That environment no longer exists. The cost of producing a plausible video of a strike that did not happen, or a plausible image of an event that happened differently, has dropped by roughly two orders of magnitude since the Protocol was drafted. The defensive implication is uncomfortable: as fabrication gets cheaper, defense counsel can credibly raise reasonable doubt about *authentic* material simply by gesturing at the broader unreliability of the evidentiary environment. Lisa Parks, writing about satellite imagery as evidence, argues that remote-sensing material draws much of its persuasive power from its appearance of objectivity, and that this appearance is fragile precisely because viewers cannot independently audit the production chain. The argument generalizes. The legitimacy of digital evidence in a tribunal setting depends on the audience’s trust in the production chain, and that trust is eroding faster than the institutional response can keep up.60
Two responses are plausible, neither yet institutionalized at the scale the situation requires. The first is cryptographic: provenance authentication at the moment of capture, through tools like the Content Authenticity Initiative’s C2PA standard, which embeds verifiable metadata in media files at the device level. Done well, this would shift the authentication burden onto whoever challenges the file, since the cryptographic signature would establish device-level provenance up front. The second is institutional: dedicated digital-forensic capacity inside the prosecuting authorities themselves, so that authentication does not depend on ad hoc collaborations with academic labs whose continued existence is not guaranteed. For Mariupol, neither is fully in place. The C2PA standard is not yet widely adopted, and most of the digital material relevant to the siege predates any cryptographic provenance infrastructure. The institutional side is also incomplete: the ICC has a digital-forensics function in the Office of the Prosecutor, but its capacity is dwarfed by the volume of material that needs assessment. The architecture is, at present, leaning on borrowed capacity from external partners, and that capacity is contingent.63
B. The Verification Gap: A Short Case Analysis
The methodological argument in this paper is that satellite-derived documentation, however rigorous, cannot stand alone as the evidentiary foundation for prosecutions at the scale Mariupol will eventually require. Defending that argument empirically is hard, because the satellite chronologies underlying Mariupol prosecutions are not publicly available at facility-level granularity. What is available is the methodological apparatus that Yale’s Humanitarian Research Lab has chosen to publish, across three reports issued between May and August 2022, and that apparatus itself supplies the empirical foundation the methodological argument needs. The case analysis that follows is small and self-limiting. It does not compare satellite data to survivor testimony — that comparison cannot be made from public sources alone. It does something more modest, and in some ways more useful: it shows, on Yale’s own published evidence, that the institution producing the most rigorous open-source documentation of Mariupol has itself, in three distinct ways, signaled the limits of what satellite-derived evidence can do.
The first signal is in Yale HRL’s Sievierodonetsk damage assessment of 29 June 2022, which catalogues damage to 51 of 89 healthcare and educational facilities in the raion between 24 February and 13 June 2022. The report includes worked examples of the satellite chronology its analysts use to identify damage. One worked example, presented in the report’s Figure 1, shows the same facility on three dates: 29 March 2022 (no visible damage), 16 May 2022 (partial damage), and 30 May 2022 (additional partial damage). The interval between the first two images is 48 days. The Yale analysts explicitly note what the methodological consequence is. “It is not possible to conclude exactly how many separate instances of damage occurred to a facility given two images separated by an extended time period,” the report states; “the chronological analysis of available imagery for the reporting period indicates the minimum number of occasions on which an identified location incurred damage.” That is the formal admission that the satellite-derived chronology, even at the level of rigor Yale exercises, undercounts the number of distinct attacks on a given facility whenever the strike rate exceeds the satellite revisit cadence. For command-responsibility prosecutions, which turn in part on whether the accused failed to act between distinguishable waves of attack, this undercount is not a minor footnote. It is the structural reason testimonial corroboration is needed to disaggregate what overhead imagery aggregates.66
The second signal is in Yale HRL’s 17 May 2022 summary on the bombardment of Ukrainian healthcare facilities, which presents Mariupol findings in aggregate form: 22 of 277 facilities verified as damaged across five cities (Mariupol, Kharkiv, Kyiv Oblast, Chernihiv, and Izyum), with 7 of those facilities in Mariupol cross-corroborated by both satellite imagery and open-source material. The summary is explicit about why the underlying facility-level data is not public. “Due to the report’s granularity and the potential use of sensitive information for targeting purposes,” the summary states, “only this summary and examples of annotated imagery depicting damage to seven (7) of these health facilities…are being released publicly. The full report will be made available as needed to international accountability mechanisms.” The redaction is methodologically defensible, since publishing facility-level coordinates would supply targeting information, but it has an evidentiary consequence that has been underappreciated in the literature. A defense lawyer, scholar, or independent expert cannot, from public sources, audit the chain of inferences linking Mariupol satellite observations to the broader claim of “widespread and systematic bombardment.” The data exists; it is structurally unavailable for independent cross-validation. For prosecutions at the conviction stage, where adversarial scrutiny of the evidentiary chain is the norm rather than the exception, this creates a vulnerability that can only be mitigated through institutional disclosure protocols specifically designed for accountability proceedings.67
The third signal, and the most important one for this paper’s methodological argument, is in Yale HRL’s 25 August 2022 report on the filtration system in Donetsk oblast. The report identifies twenty-one distinct filtration locations and maps the procedural categories that operate at each: registration, holding, secondary interrogation, release, transport to Russia, and short- or long-term detention. The agencies reporting filtration figures are explicit in the report: the former Ukrainian Ombudswoman for Human Rights gave a figure of 1,700,000 civilians transferred from Ukraine to Russia by 14 June 2022, including 276,000 children; the Russian government, on 16 August 2022, claimed twice that number — 3,400,000, including over 550,000 children. The contested scale of the system is itself an evidentiary fact, and one that satellite imagery cannot resolve. What Yale states in its own report is the methodological point that matters most: “The total number of facilities engaged in filtration operations in Donetsk oblast cannot be conclusively determined by open source analysis and remote sensing alone. Ground investigations are required to cross-corroborate these initial findings.” This is not a critic’s argument about Yale’s methodology. It is Yale’s own statement, embedded in the foundational public assessment of the filtration system, that satellite-derived documentation has structural limits that only ground-level, testimonial work can close. The methodological argument of this paper aligns with that statement: the testimonial layer is not supplemental to the satellite layer; it is the only mechanism by which certain classes of evidentiary claims can be made at all.68
Three things follow from this short analysis. First, that the chronological undercount built into satellite revisit cadences is a known methodological limit, acknowledged by the institution that has done the most rigorous published work on Mariupol. Second, that facility-level Mariupol satellite chronologies, while they exist within Yale’s institutional archive, are not publicly available for cross-validation, which means that defense-side scrutiny at trial will have to rely on disclosure mechanisms that have not yet been formally designed. Third, that the filtration system, the most consequential subset of the Mariupol record for crimes-against-humanity and possible genocide prosecutions, is, on Yale’s own assessment, partially inaccessible to satellite documentation, and structurally requires ground-level corroboration. These three findings do not require survivor testimony to establish. They follow from a close reading of the methodological documents Yale has chosen to publish. They support the broader methodological argument of this paper not by analogy or by appeal to lived experience, but by direct citation to the published work of the institution whose satellite documentation underlies the ICC arrest warrants currently in force against Russian leadership.
A critic might argue that this reading overstates what Yale’s methodological caveats actually concede. The undercount language in the Sievierodonetsk report, on this reading, is standard methodological hedging; every empirical study includes such caveats, and treating them as structural admissions misreads the genre. Similarly, the redaction language in the May 2022 healthcare summary protects sources rather than concealing weakness, and the “ground investigations are required” phrase in the filtration report is forward-looking research planning, not a limitation on the published findings. Each of those readings has some merit. The genre of methodological caveats does flatten the difference between routine hedging and structural admission, and I am reading these passages closely enough that other readings remain open. But the cumulative pattern matters. A single caveat could be discounted as boilerplate. Three caveats, across three reports, each pointing at the same methodological boundary, the limit of what remote sensing can do without ground-level corroboration, are not boilerplate. They are a consistent institutional position, and the fact that Yale has chosen to publish that position in three different formal documents is itself the evidence. My reading is not that Yale’s reports are weak; it is that they are exactly as strong as they claim to be, and that what they claim falls short of what prosecutions at the conviction stage will require.
A fourth signal points in the same methodological direction but from the opposite side. The 2024 Human Rights Watch and Truth Hounds report on Mariupol, prepared with the visual investigations practice SITU Research, is the most thorough published attempt to integrate satellite imagery, open-source material, and survivor testimony into one evidentiary architecture for the siege. The report draws on 240 interviews, more than 850 verified photographs and videos, dozens of satellite images, and 3D photogrammetric reconstructions of seven buildings. Yale’s reports map the outer edge of what satellite-only work can do. The HRW report maps what becomes possible once a testimonial layer is built in at scale. This is not a critique of either institution — the two operate under different mandates, with different procedural constraints, and produce different outputs because of it. But put side by side, they make a single methodological point that neither makes on its own.69
The report’s estimate of the dead is where this shows most sharply. By analysing very-high-resolution satellite imagery and drone footage of five cemeteries in and around Mariupol, counting both individual graves visible in overhead imagery and the smaller plaques and crosses that mark trench-style mass graves, which are only visible from the ground, HRW estimates that at least 10,284 people were buried in the city between March 2022 and February 2023, of whom roughly 8,000 likely died from war-related causes above the baseline mortality rate for a city of that size. The same report notes that Mariupol’s then-mayor estimated 22,000 dead, and that the UN Monitoring Mission’s recorded civilian deaths stood at fewer than 2,000. The three figures aren’t contradictory; they measure different things, and the gap between them is the epistemological distance this paper is about. The UN figure is what could be independently verified by an external monitoring body without access to the city. The mayoral estimate is what local officials, drawing on testimonial channels and a fragmented infrastructure of municipal record-keeping, believe to be true. The HRW figure sits in the middle: rigorous enough to be defended in legal proceedings, but only reachable because the satellite layer was integrated with ground-level knowledge of how the cemeteries were actually being used. A satellite alone counts gravesites. It doesn’t know what is in them, and it doesn’t know what has been moved.
The case studies in the report make the same point at the level of individual incidents. The HRW reconstruction of the attack on the apartment building at Mytropolytska Street 98, for example, combines satellite chronology of the structural damage with testimony from a survivor who was in an adjacent basement room at the moment of the strike, accounts from members of the recovery teams who worked the site between April 21 and April 25, and the testimony of a relative of victims who returned to the building repeatedly between April 27 and September 18 looking for remains. The satellite record establishes when the building was struck and the spatial extent of the collapse. The testimonial layer establishes how many people were in which basement rooms at the moment of impact, who recovered the bodies, what condition the bodies were in, what was done with them afterwards, and whose names belong to which set of remains. Both layers are necessary to the evidentiary picture. Neither, on its own, supports a prosecution against the individual or chain of command responsible. The argument of this paper, that a testimonial layer is structurally required rather than optional, finds in the HRW Mytropolytska reconstruction a worked example on the record.
The four signals converge on a single claim. Satellite-based documentation, even at the rigor Yale has achieved, runs into structural limits that its own publications acknowledge. Closing those limits requires a testimonial architecture of the kind HRW and Truth Hounds have shown is possible to build, even under occupation, if the methodological work is treated as primary rather than as a supplement. The institutional implication, picked up in Part VII, is that future Mariupol prosecutions will only reach their full evidentiary potential if both layers are funded, protected, and harmonized as a single integrated system, instead of being treated as the products of two unrelated workflows that occasionally cite one another.
C. The Witness Protection Problem
A third structural problem concerns witnesses. Mariupol generated an enormous pool of potential witnesses: survivors who fled the city, soldiers who participated in its defense, civilians who were processed through filtration sites and then released. Each category presents its own witness-protection challenges. Survivors are scattered across Europe, North America, and Russian-controlled territory, and the conditions under which their initial testimonies were taken vary enormously. Soldiers face operational-security issues and the risk of reciprocal prosecution if they enter Russian territory. Filtration survivors often still have family members in Russian-controlled areas, leaving them open to coercion and retaliation. International criminal tribunals have built up substantial witness-protection capacity, but the scale of the Ukraine situation strains the existing systems. The ICC’s Victims and Witnesses Unit, the ICTY/ICTR experience with witness intimidation, and the more recent practice of remote testimony all offer partial templates, but none scales to the volume of potential witnesses the Mariupol case alone could generate. The aggregate-corroboration logic that arguably saves the digital-evidence problem cuts the other way for witnesses: when individual testimony cannot be fully protected, tribunals have historically leaned more heavily on convergent independent accounts. Doing so at scale requires both protection infrastructure and an evidentiary procedure that can metabolize hundreds or thousands of partial witnesses without collapsing into anonymity.
D. In Absentia Proceedings and Immunities
A fourth structural problem is procedural. Russia is unlikely to surrender any of its leadership to international or national tribunals, which raises two linked issues. The first is in absentia proceedings. Article 63(1) of the Rome Statute provides that “[t]he accused shall be present during the trial,” subject to narrow exceptions for disruptive behavior. The Statute of the Special Tribunal for the Crime of Aggression against Ukraine takes a more flexible approach, authorizing investigation and indictment to proceed in the accused’s absence, with full trial proceedings to follow once the accused is present. National jurisdictions vary: the French model is comparatively permissive of in absentia trial, the German one allows it only in narrow circumstances.47
The fair-trial constraints on in absentia proceedings are real. Article 14(3)(d) of the International Covenant on Civil and Political Rights guarantees the accused the right “[t]o be tried in his presence.” The European Court of Human Rights, in Sejdovic v. Italy (Grand Chamber, 2006), held that an in absentia conviction must be followed by a right to retrial in the accused’s presence, unless there is evidence that the accused has unequivocally waived the right to be present.48
The second issue is immunities. Heads of state, heads of government, and foreign ministers, the so-called troika, enjoy personal immunity (immunity ratione personae) in foreign courts under customary international law while in office. Article 27 of the Rome Statute partially abrogates this immunity, providing that official capacity is no bar to prosecution; whether Article 27 binds non-party states remains contested. The ICC’s Al-Bashir Jordan judgment took the position that no head-of-state immunity exists vis-à-vis an international court; the French Court of Cassation’s July 2025 ruling in the al-Assad case illustrates the depth of the disagreement. The Special Tribunal Statute aligns with Al-Bashir. Whether that position survives challenge in national courts remains to be tested.50
E. The Time Problem
A fifth structural problem, and perhaps the most underappreciated, is time. International criminal proceedings move on a timeline fundamentally mismatched to the timeline of political memory. The ICC investigations into Darfur, Libya, and Georgia have stretched across more than a decade. The ad hoc tribunals for the former Yugoslavia and Rwanda each ran for more than twenty years, with the ICTY closing in December 2017 and the ICTR in December 2015. The Special Tribunal for the Crime of Aggression against Ukraine, signed into existence in June 2025, was still in the establishment phase in mid-2026. The evidence architecture has to be built to outlast not only the war, but also the political conditions that produced the will to document it.52
VII. Evidence as Strategic Infrastructure: A Policy Framework
The argument to this point has been that the gap between documentation and adjudication is not a problem of substantive law. The law is largely settled. The gap lies in the institutional architecture that connects the two, and in the doctrinal disputes that decide what gets proved. Whether that architecture can survive the timeline international prosecutions actually require, and the political conditions it will actually have to operate under, is the real question. What follows is a four-part sketch of a policy framework that takes evidence preservation seriously as infrastructure, on a par with deterrence and reconstruction rather than apart from them.
A. Long-Horizon Institutional Funding
Begin with funding. Evidence-collection institutions cannot run on annual appropriations when the prosecutions they feed take a decade or longer to mature. Yale’s Humanitarian Research Lab came close to closure in 2025 after federal funding was interrupted; donor support later kept the Lab operating through at least October 1, 2025, but the episode still exposed how fragile the architecture is. That is not a sustainable model for an institution whose work has contributed to accountability efforts involving a sitting head of state. The minimum fix is structural: multi-year, treaty-anchored pooled funding, drawn from Rome Statute parties, Council of Europe Enlarged Partial Agreement signatories, and any third states with material interest in the Ukraine accountability process. The ICPA is already funded that way. The institutional model is not novel: the International, Impartial and Independent Mechanism on Syria (IIIM), established by UN General Assembly Resolution 71/248 in 2016, runs on voluntary contributions pooled through the UN, with funding cycles long enough to support work that extends beyond any single national budget cycle. That model is imperfect — it remains vulnerable to large-donor withdrawal — but it is materially more robust than the annual-appropriations posture under which much current documentation work operates. Extending the IIIM funding template to academic and civil-society documentation work that already feeds international prosecutions would close perhaps the single most consequential vulnerability in the architecture as it stands.65
B. Harmonization of Evidentiary Standards
Funding is half of the problem. The other half is interoperability. Evidence collected by Yale’s satellite work feeds into ICC investigations, into OSCE fact-finding, into German universal-jurisdiction prosecutions, into the Register of Damage, and into the Special Tribunal. Each institution has its own procedural requirements, but the underlying evidentiary standards should not vary so widely that material gathered for one body cannot be used by another. The Berkeley Protocol on Digital Open Source Investigations provides a baseline for digital material. Nothing comparable yet exists for witness testimony, for forensic reconstruction, or for the integration of structural and case-specific investigation. Bellingcat’s own methodology has become a de facto standard in open-source work, which is useful but informal. A more formal successor to the Berkeley Protocol, focused on conflict-specific evidentiary standards across institutions, would strengthen the system considerably without forcing any institution to surrender its own procedural autonomy. One might object that harmonization at this level is a solution in search of a problem — that the existing institutional ecosystem, with its productive redundancy, already produces usable cross-jurisdictional evidence. That objection has some force. Redundancy is not the same as fragmentation, and ad hoc translation between institutional formats has worked, so far, in lower-volume contexts. But the Ukraine situation is not a lower-volume context, and at its actual scale, the marginal cost of ad hoc translation rises faster than the marginal benefit of preserving institutional independence. Harmonization here is not about imposing a single standard; it is about establishing a translation protocol between standards, so that material produced under one regime is portable to another without re-authentication from scratch.
Concretely, this would mean a working group co-convened by the ICC’s Office of the Prosecutor and the ICPA, with formal participation by the German Federal Public Prosecutor’s Office, the French Crimes Against Humanity Unit, the Special Tribunal Registry once established, and Eurojust’s JIT secretariat. Its first deliverable would be a cross-jurisdictional admissibility matrix mapping, for each major evidentiary category — satellite imagery with chain-of-custody documentation, geolocated open-source video, intercepted communications, witness testimony with associated procedural conditions, and forensic reconstruction outputs — what each institution requires for admissibility, what counts as equivalent, and what gaps must be filled by re-authentication. The matrix would not impose a single standard; it would simply make the existing standards legible to one another. A second deliverable would be a translation protocol for chain-of-custody documentation, so that material gathered under, say, the German structural-investigation procedure can be presented to the Special Tribunal without the metadata loss that currently characterizes ad hoc institutional transfers. None of this requires treaty amendment, none of this overrides institutional procedural autonomy, and all of it is achievable on a two-to-three-year timeline if the relevant institutions can agree on the working-group’s convening authority. That last condition is the binding one. Without sustained political pressure from State Parties, no institution will surrender the convening authority it already exercises. The IIIM funding template proposed above could be paired with a small dedicated harmonization secretariat to provide that convening function.58
C. The Local Knowledge Layer
A third issue is methodological. Even rigorous institutions like Yale’s HRL depend on local knowledge, and that knowledge is a depleting resource. The cohort of people who lived through the Mariupol siege and can speak authoritatively to what happened, those who knew the city itself and not only what the satellite recorded, is finite in number. Some are dead. Some remain inside the occupied city. Others were transferred to Russia and have not returned. Those who left are dispersed across Europe and North America. What this cohort holds (the city’s topography, the patterns of daily life under siege, the chronology at neighborhood scale, the social context of filtration) is not recoverable from imagery or open-source social media. The methodological note at the opening of this paper is relevant in this connection: the author’s family is from Mariupol. Lived proximity does not adjudicate questions of law, but it does provide a vantage point from which the discrepancy between what satellites recorded and what the city actually was becomes legible. Closing that discrepancy is among the more practical tasks facing the Ukraine evidence architecture. It is also a condition of legitimacy: international criminal law is judged not only by legal correctness, but by whether the verdicts it produces are recognized as accurate by those most directly affected.53
Consider what satellite imagery cannot capture, and what only local knowledge can fill in. Three examples.
First, the collapse of the cellular grid and the emergence of informal communication networks. In the western districts of Mariupol, cellular service was effectively gone by the fourth day of the invasion; electricity followed within forty-eight hours. The institutional record captures these collapses as undifferentiated outages. What it does not capture is the architecture that filled the void. Building stairwells became message boards. People with charged phones moved between buildings on foot, carrying news. A handful of car radios in working vehicles became collective information sources. Information moved by the kilometer per day, not the second. This matters in two specific ways. It shapes witness reliability years later. A survivor recalling the sequence of events in March 2022 is reconstructing a chronology that was already, at the time, assembled from fragments rather than experienced as continuous. The chronological errors a defense lawyer can exploit are not errors of memory in the usual sense; they are the artifact of an information environment that never produced a continuous timeline in the first place. And it shapes the institutional record itself: the absence of cellular metadata for the first two weeks of the siege is not a gap to be lamented but an evidentiary fact in its own right, with implications for what kinds of corroboration a tribunal can reasonably demand.64
Second, the divergence between satellite chronology and lived chronology at neighborhood scale. Yale’s study of damage to medical facilities works at a city-wide aggregate level, and at that level the dataset is rigorous. But the strike chronology satellite imagery yields (a pre-strike image, a post-strike image, with the interval set by the satellite’s revisit cadence) tends to blur sequential events into single timepoints. What survivors in the western districts will describe as a single morning of bombardment frequently shows up in satellite chronology as a damage state observed over a 48-to-72-hour interval. The legal stakes are not trivial. Command-responsibility cases turn in part on whether a senior commander could have intervened between an initial wave of strikes and a subsequent one. A satellite chronology that compresses two waves into a single observation interval supplies weaker evidence for the “failed to prevent or repress” element than a lived chronology that distinguishes them. The remedy is not to throw out the satellite data, but to integrate it with neighborhood-scale testimonial material that can disaggregate what overhead imagery aggregates.
Third, the social structure of filtration. Yale HRL’s 2022 report mapped twenty-one sites and the procedural categories that operated at each: registration, holding, secondary interrogation. What the satellite record cannot show is the line between filtration sites that processed civilians on what was essentially an administrative basis and sites that functioned, at certain points, as instruments of selective extraction. The legal distinction matters. An administratively organized filtration system, even one that produced torture and abuse, is doctrinally different from one in which certain populations (men of military age, public servants, identifiable Ukrainian-language speakers) were systematically separated for distinct treatment. The first set of facts supports war-crimes charges. The second substantially strengthens an argument for crimes against humanity, and possibly for the population-destruction element of genocide. The empirical line between the two cannot be drawn from satellite imagery alone. It requires testimonial work organized around what survivors of specific filtration sites can describe about who was held, who was released, who was not, and on what apparent criteria. This is not material the institutional record can produce on its own. Veena Das has written, in a different context, about the gap between event-as-documented and event-as-lived; the gap in the Mariupol filtration record is the kind she has in mind, and closing it is among the more pressing tasks of any serious testimonial collection effort.61
A future tribunal that reconstructs Mariupol only from remote-sensing data will produce an accurate but partial city. A tribunal that integrates testimonial material gathered under proper procedural safeguards will produce a city the survivors can recognize. The second standard is the one an evidence architecture ought to be aiming for, and the practical work of meeting it is methodological, not aspirational. It requires building structured testimonial collection into the documentation pipeline at the same level of formality as satellite analysis, with the same chain-of-custody discipline, the same procedural protocols, and the same long-horizon funding. Bellingcat’s open-source methodology, the Berkeley Protocol’s digital-evidence standards, and Yale’s remote-sensing pipeline have analogues; structured testimonial integration does not yet have one. Developing it is the most consequential methodological task the next phase of the evidence architecture should set itself.
D. Connecting Accountability to Reconstruction
Finally, accountability and reconstruction. The donor-state reflex has been to treat the two as separate streams: documentation now, reconstruction later. The division is artificial. Decisions about how Mariupol is rebuilt are themselves evidentiary decisions: whether to preserve the Drama Theater ruins as a memorial, whether to demolish or restore Azovstal, how the names of the missing appear (or do not) in the new cityscape. They determine what the city looks like to a tribunal in 2035 or to a historian in 2070. The transitional-justice and post-conflict-memory literature, developed after the Balkans, Rwanda, and elsewhere, has been consistent on this point: physical reconstruction of contested spaces is part of accountability, not parallel to it. The practical implication for Mariupol is that reconstruction planning needs to treat evidentiary and memorial sites as a non-negotiable category, alongside utilities and housing rather than after them.59
VIII. Conclusion
Azovstal’s last defenders surrendered in May 2022. The city was destroyed. The siege was over. The legal architecture was only beginning. Four years on, that architecture is more developed than it has been at any prior moment in international criminal law. Six ICC arrest warrants, the first of them issued against a sitting head of state of a permanent member of the UN Security Council. The methodological standards Yale’s HRL and the Conflict Observatory have set for academic atrocity documentation. Cross-border prosecutorial coordination on a scale never tried before, through the Joint Investigation Team and the ICPA. A Special Tribunal for Aggression, signed into existence in June 2025 to close the most consequential jurisdictional gap in the Rome Statute regime. A civil-reparations pathway through the Council of Europe Register of Damage and the International Claims Commission that has no real precedent since 1945.
And yet what comes next for Mariupol legally is uncertain. The architecture is necessary, but nowhere near sufficient. It can be defunded: the Yale crisis in 2025 was a preview of how easily. It can be politically captured. It can be outpaced by the rate at which political memory decays. It can fracture along the joints: between criminal prosecution and civil reparation, between The Hague and the national courts, between the ICC and the Special Tribunal, between the documentary record and the testimonial one. And it has to function within doctrinal disputes that are not resolved: the Bemba narrowing of command responsibility, the Al-Bashir controversy over head-of-state immunity, the Al-Werfalli precedent for digital evidence that has not yet been tested at the conviction stage. Any of these fractures, if it widens, leaves Mariupol where so many earlier atrocities ended up: well documented, never adjudicated.
The argument of this paper has been that the gap between documentation and adjudication is closable, but only if evidence preservation is treated as strategic infrastructure rather than humanitarian afterthought. The specific institutional implications follow from that framing: long-horizon funding on the IIIM template rather than the annual-appropriations model; harmonization protocols between institutional evidentiary standards rather than ad hoc translation; structured testimonial integration developed to the same level of formality as digital-evidence collection; reconstruction planning that treats memorial and evidentiary sites as a non-negotiable category. None of this is radical. Each piece builds on existing practice. What it requires is the recognition that whatever comes next in the European security order will be shaped, in significant measure, by whether the violence that did this to a city like Mariupol can reliably be translated into legal record.
Here is my forecast, not a legal inevitability. The cautious nothing-conclusion academic papers tend to default to would be easier, but Mariupol does not deserve that. In the medium term, over the next five to seven years, the Mariupol record is likely to be tested at the conviction stage in at least one of three forums: a Special Tribunal proceeding against a member of the Russian political leadership; a national universal-jurisdiction proceeding in Germany or France against a mid-level commander; or, more remotely, an ICC trial following the surrender or capture of one of the individuals already named in arrest warrants. The Special Tribunal route is the most politically charged and the most doctrinally fragile, given the immunity questions canvassed in Part V. The national route is the most likely to produce an actual conviction but the least likely to produce a verdict that reaches the political leadership directly responsible. The ICC route depends on contingencies outside the legal system altogether. Whichever of these proceedings reaches conviction first will determine, more than any scholarly debate, what the Bemba, Al-Bashir, and Al-Werfalli precedents actually mean for Mariupol. The structural argument of this paper, that evidence preservation needs to be funded, harmonized, and methodologically extended to local-knowledge work, is meant to ensure that whichever forum gets there first does so on the strongest evidentiary footing available. The siege ended in May 2022. The legal proceedings that follow from it have a longer arc than the war. What happens in the next decade will determine whether Mariupol becomes the case that closes the gap between documentation and adjudication, or the case that finally demonstrates the gap cannot be closed.
My personal stake in the outcome is not, I think, what should decide the answer. The methodological note at the start of this paper was meant to make clear that lived proximity to the siege does not adjudicate questions of law. Proximity does, however, sharpen the question the whole architecture is meant to answer: whether the international legal system can produce a verdict the people who lived through what happened in Mariupol can recognize as accurate. That is a higher standard than admissibility. It is the standard this paper has tried to meet.
Selected Bibliography
Treaties and Legal Instruments
Agreement between the Council of Europe and Ukraine on the Establishment of the Special Tribunal for the Crime of Aggression against Ukraine. Strasbourg, June 25, 2025.
Convention on the Prevention and Punishment of the Crime of Genocide. December 9, 1948. 78 U.N.T.S. 277.
Geneva Convention Relative to the Protection of Civilian Persons in Time of War (Fourth Geneva Convention). August 12, 1949. 75 U.N.T.S. 287.
International Covenant on Civil and Political Rights. December 16, 1966. 999 U.N.T.S. 171.
Rome Statute of the International Criminal Court. July 17, 1998. 2187 U.N.T.S. 90.
Statute of the Special Tribunal for the Crime of Aggression against Ukraine. Annexed to the Council of Europe–Ukraine Agreement of June 25, 2025.
Case Law
Prosecutor v. Al-Bashir, ICC-02/05-01/09-397. Judgment in the Jordan Referral re Al-Bashir Appeal. Appeals Chamber, International Criminal Court, May 6, 2019.
Prosecutor v. Akayesu, Case No. ICTR-96-4-T. Judgment. International Criminal Tribunal for Rwanda, September 2, 1998.
Prosecutor v. Al-Werfalli, ICC-01/11-01/17. Warrant of Arrest. International Criminal Court, August 15, 2017; Second Warrant of Arrest, July 4, 2018.
Prosecutor v. Bemba Gombo, ICC-01/05-01/08. Judgment Pursuant to Article 74 of the Statute. Trial Chamber III, International Criminal Court, March 21, 2016.
Prosecutor v. Bemba Gombo, ICC-01/05-01/08 A. Judgment on the Appeal. Appeals Chamber, International Criminal Court, June 8, 2018.
Prosecutor v. Krstić, Case No. IT-98-33-A. Judgment. Appeals Chamber, International Criminal Tribunal for the Former Yugoslavia, April 19, 2004.
Prosecutor v. Tadić, Case No. IT-94-1-AR72. Decision on the Defence Motion for Interlocutory Appeal on Jurisdiction. International Criminal Tribunal for the Former Yugoslavia, October 2, 1995.
Sejdovic v. Italy, Application No. 56581/00. European Court of Human Rights, Grand Chamber, March 1, 2006.
Reports and Institutional Documents
Amnesty International. "Russia's Attack on the Mariupol Drama Theatre Was a War Crime." June 30, 2022.
Council of Europe. Register of Damage Caused by the Aggression of the Russian Federation against Ukraine: Statute. Committee of Ministers Resolution CM/Res(2023)3, May 17, 2023; Register of Damage for Ukraine, Annual Activity Report 2025; Council of Europe reporting on submitted and recorded claims, April 30, 2026.
International Criminal Court. "Situation in Ukraine: ICC Judges Issue Arrest Warrants Against Vladimir Vladimirovich Putin and Maria Alekseyevna Lvova-Belova." March 17, 2023.
OSCE Office for Democratic Institutions and Human Rights. Report on Violations of International Humanitarian and Human Rights Law, War Crimes and Crimes Against Humanity Committed in Ukraine since 24 February 2022. Warsaw: OSCE/ODIHR, April 2022.
Parliamentary Assembly of the Council of Europe. Resolution 2495 (2023): Legal and Human Rights Aspects of the Russian Federation's Aggression against Ukraine.
United Nations Independent International Commission of Inquiry on Ukraine. Report A/HRC/52/62. March 15, 2023.
United Nations Office of the High Commissioner for Human Rights. Berkeley Protocol on Digital Open Source Investigations. Geneva and Berkeley, 2022.
Yale School of Public Health, Humanitarian Research Lab. Russia's Systematic Program for the Re-education and Adoption of Ukraine's Children. New Haven: Conflict Observatory, 2023. See also Children of War platform, official consolidated data on Ukrainian children affected by Russia's war, data as of May 31, 2026, https://childrenofwar.gov.ua/en/.
Yale School of Public Health, Humanitarian Research Lab. System of Filtration: Mapping Russia's Detention Operations in Donetsk Oblast. New Haven: Conflict Observatory, August 25, 2022.
Khoshnood, Kaveh, and Nathaniel A. Raymond et al. "Damage Assessment of Health and Educational Facilities in Sievierodonetsk Raion, Ukraine: Evidence of Widespread, Indiscriminate, and Persistent Bombardment by Russia and Russia-Aligned Forces Between 24 February - 13 June 2022." Yale School of Public Health, Humanitarian Research Lab. New Haven: Conflict Observatory, June 29, 2022.
Khoshnood, Kaveh, Nathaniel A. Raymond et al. "Evidence of Widespread and Systematic Bombardment of Ukrainian Healthcare Facilities: Summary." Yale School of Public Health, Humanitarian Research Lab. New Haven: Conflict Observatory, May 17, 2022.
Scholarly Sources
Amann, Diane Marie. "In Bemba and Beyond, Crimes Adjudged to Commit Themselves." EJIL: Talk!, June 13, 2018.
Poole, Danielle N., Oona A. Hathaway, Daniel Andersen, Nathaniel A. Raymond, Jack Parham, Caitlin N. Howarth, and Kaveh Khoshnood. "The Effect of Conflict on Damage to Medical Facilities in Mariupol, Ukraine: A Quasi-Experimental Study." PLOS Global Public Health, January 2025.
Bassiouni, M. Cherif. Introduction to International Criminal Law. 2nd revised edition. Leiden: Martinus Nijhoff Publishers, 2013.
Cassese, Antonio. International Criminal Law. 3rd edition. Oxford: Oxford University Press, 2013.
Dannenbaum, Tom. The Crime of Aggression, Humanity, and the Soldier. Cambridge: Cambridge University Press, 2018.
Dannenbaum, Tom. "Mechanisms for Criminal Prosecution of Russia's Aggression Against Ukraine." Just Security, March 10, 2022.
Das, Veena. Life and Words: Violence and the Descent into the Ordinary. Berkeley: University of California Press, 2007.
David, Lea. The Past Can't Heal Us: The Dangers of Mandating Memory in the Name of Human Rights. Cambridge: Cambridge University Press, 2020.
DeGuzman, Margaret M. "Gravity and the Legitimacy of the International Criminal Court." Fordham International Law Journal (2009).
Freeman, Lindsay. "Digital Evidence and War Crimes Prosecutions: The Impact of Digital Technologies on International Criminal Investigations and Trials." Fordham International Law Journal (2018).
Hamilton, Rebecca J. "The Hidden Costs of Open-Source Evidence." Texas National Security Review, Fall 2023.
Hnatovskyy, Mykola. "Міжнародне гуманітарне право та збройний конфлікт в Україні" [International Humanitarian Law and the Armed Conflict in Ukraine]. In Український щорічник міжнародного права [Ukrainian Yearbook of International Law]. Kyiv, 2023.
Human Rights Watch and Truth Hounds. "Our City Was Gone: Russia's Devastation of Mariupol, Ukraine." With 3D reconstructions and visual and spatial analysis by SITU Research. New York: Human Rights Watch, 2024.
Koenig, Alexa. Scholarship on open-source investigations and digital evidence, Human Rights Center, UC Berkeley School of Law.
McDougall, Carrie. The Crime of Aggression Under the Rome Statute of the International Criminal Court. 2nd edition. Cambridge: Cambridge University Press, 2021.
Olick, Jeffrey K. The Politics of Regret: On Collective Memory and Historical Responsibility. New York: Routledge, 2007.
Parks, Lisa. Cultures in Orbit: Satellites and the Televisual. Durham: Duke University Press, 2005.
Parks, Lisa. "Vertical Mediation and the U.S. Drone War in the Horn of Africa." In Life in the Age of Drone Warfare, edited by Lisa Parks and Caren Kaplan. Durham: Duke University Press, 2017.
Osiel, Mark. Mass Atrocity, Collective Memory, and the Law. New Brunswick: Transaction Publishers, 1997.
Ritscher, Christian. "COVID-19 and International Crimes Trials in Germany." 18 Journal of International Criminal Justice 1077 (2020).
SáCouto, Susana, and Patricia Sellers. "The Bemba Appeals Judgment: Impunity for Sexual and Gender-Based Crimes." 27 William & Mary Bill of Rights Journal 599 (2019).
Sadat, Leila N. "Prosecutor v. Jean-Pierre Bemba Gombo." 113 American Journal of International Law 353 (2019).
Safferling, Christoph. International Criminal Procedure. Oxford: Oxford University Press, 2012.
Schabas, William A. An Introduction to the International Criminal Court. 6th edition. Cambridge: Cambridge University Press, 2020.
Subotić, Jelena. Yellow Star, Red Star: Holocaust Remembrance after Communism. Ithaca: Cornell University Press, 2019.
Van Schaack, Beth, and Tom Dannenbaum. "Building a Special Tribunal for the Crime of Aggression against Ukraine." EJIL: Talk!, July 25, 2025.
Notes
- 1
Office of the High Commissioner for Human Rights, Report on the Human Rights Situation in Ukraine, 1 February – 31 July 2022 (Geneva: OHCHR, 2022) ¶¶ 27–42, characterizing the destruction of Mariupol as among the most severe in any European urban area since 1945.
↩ - 2
International Criminal Court, "Statement of ICC Prosecutor Karim A.A. Khan QC on the Situation in Ukraine: Receipt of Referrals from 39 States Parties and the Opening of an Investigation," March 2, 2022, https://www.icc-cpi.int/news/statement-icc-prosecutor-karim-aa-khan-qc-situation-ukraine-receipt-referrals-39-states.
↩ - 3
Yale School of Public Health, Humanitarian Research Lab, Russia's Systematic Program for the Re-education and Adoption of Ukraine's Children (New Haven: Conflict Observatory, 2023), identifying at least 6,000 Ukrainian children and at least forty-three facilities. For current official Ukrainian figures on deported and/or forcibly displaced children and returned children, see Children of War platform, data as of May 31, 2026, https://childrenofwar.gov.ua/en/.
↩ - 4
Yale Humanitarian Research Lab, "System of Filtration: Mapping Russia's Detention Operations in Donetsk Oblast," August 25, 2022, https://hub.conflictobservatory.org/.
↩ - 5
Council of Europe and Ukraine, Agreement on the Establishment of the Special Tribunal for the Crime of Aggression against Ukraine, signed June 25, 2025, Strasbourg, art. 1 (jurisdiction); Statute of the Special Tribunal, annexed thereto.
↩ - 6
International Criminal Court, "Situation in Ukraine: ICC Judges Issue Arrest Warrants Against Vladimir Vladimirovich Putin and Maria Alekseyevna Lvova-Belova," March 17, 2023.
↩ - 7
Rome Statute of the International Criminal Court, July 17, 1998, arts. 8(2)(a)(vii), 8(2)(b)(viii), 2187 U.N.T.S. 90 [hereinafter Rome Statute]. Article 8(2)(a)(vii) criminalizes unlawful deportation or transfer of protected persons, including children. Article 8(2)(b)(viii) criminalizes the transfer, directly or indirectly, by the occupying power of parts of its own civilian population into the territory it occupies, or the deportation or transfer of all or parts of the population of the occupied territory within or outside this territory.
↩ - 8
Danielle N. Poole et al., "The Effect of Conflict on Damage to Medical Facilities in Mariupol, Ukraine: A Quasi-Experimental Study," PLOS Global Public Health (January 2025), https://journals.plos.org/globalpublichealth/article?id=10.1371/journal.pgph.0003950. Coauthors include Oona A. Hathaway, Daniel Andersen, Nathaniel A. Raymond, Jack Parham, Caitlin N. Howarth, and Kaveh Khoshnood.
↩ - 9
Geneva Convention Relative to the Protection of Civilian Persons in Time of War, August 12, 1949, art. 18, 75 U.N.T.S. 287 [hereinafter Geneva Convention IV] ("Civilian hospitals organized to give care to the wounded and sick, the infirm and maternity cases, may in no circumstances be the object of attack, but shall at all times be respected and protected by the Parties to the conflict.").
↩ - 10
Rome Statute, art. 8(2)(b)(ix); see also International Committee of the Red Cross, Customary International Humanitarian Law, Rules 28 and 35, recognizing the protection of medical units and the duty of feasible precautions as customary law binding on all parties.
↩ - 11
OSCE Office for Democratic Institutions and Human Rights, Report on Violations of International Humanitarian and Human Rights Law, War Crimes and Crimes Against Humanity Committed in Ukraine since 24 February 2022 (Warsaw: OSCE/ODIHR, April 2022), invoked under the OSCE's Moscow Mechanism by forty-five participating States.
↩ - 12
Associated Press, "AP Evidence Points to 600 Dead in Mariupol Theater Airstrike," May 4, 2022.
↩ - 13
Amnesty International, "Russia's Attack on the Mariupol Drama Theatre Was a War Crime," June 30, 2022.
↩ - 14
See United Nations Independent International Commission of Inquiry on Ukraine, Report A/HRC/52/62 (March 15, 2023) ¶ 95 et seq., characterizing the forcible transfer of children as a war crime and noting the systematic and widespread nature of the deportations.
↩ - 15
Convention on the Prevention and Punishment of the Crime of Genocide, December 9, 1948, art. II(e), 78 U.N.T.S. 277 [hereinafter Genocide Convention]: genocide includes "[f]orcibly transferring children of the group to another group."
↩ - 16
Parliamentary Assembly of the Council of Europe, Resolution 2495 (2023) ("Legal and human rights aspects of the Russian Federation's aggression against Ukraine") ¶¶ 5, 12.
↩ - 17
Rome Statute, arts. 11–12. Ukraine accepted the jurisdiction of the Court for crimes committed on its territory through two declarations under Article 12(3): the first, lodged on April 17, 2014, accepted ICC jurisdiction over crimes committed in Ukrainian territory from November 21, 2013 to February 22, 2014; the second, lodged on September 8, 2015, extended the acceptance on an open-ended basis from February 20, 2014 onwards. Ukraine signed the Rome Statute on January 20, 2000, but did not ratify for nearly twenty-five years owing to a 2001 Ukrainian Constitutional Court ruling that required a constitutional amendment, which was completed in 2016 with entry into force on June 30, 2019. The Verkhovna Rada voted to ratify the Rome Statute and its amendments on August 21, 2024 (281 votes in favor); President Zelenskyy signed the ratification law on August 24, 2024; the implementing legislation amending the Ukrainian Criminal Code and Criminal Procedure Code was adopted on October 9, 2024, and signed by the President on October 22, 2024; Ukraine deposited its instrument of ratification with the United Nations Secretary-General on October 25, 2024 (UN Treaty Section, C.N.440.2024.TREATIES-XVIII.10); pursuant to Article 126 of the Rome Statute, the treaty entered into force for Ukraine on January 1, 2025, making Ukraine the 125th State Party. The pre-ratification Article 12(3) declarations remain the operative jurisdictional basis for the period addressed in this paper.
↩ - 18
See generally William A. Schabas, An Introduction to the International Criminal Court, 6th ed. (Cambridge: Cambridge University Press, 2020), on the jurisdictional architecture of the ICC and the limits imposed by the Kampala Amendments' carve-out in Article 15 bis(5).
↩ - 19
Rome Statute, art. 15 bis(5): "In respect of a State that is not a party to this Statute, the Court shall not exercise its jurisdiction over the crime of aggression when committed by that State's nationals or on its territory."
↩ - 20
Rome Statute, art. 28(a) (responsibility of military commanders): liability attaches where the commander "either knew or, owing to the circumstances at the time, should have known that the forces were committing or about to commit such crimes" and "failed to take all necessary and reasonable measures within his or her power to prevent or repress their commission or to submit the matter to the competent authorities for investigation and prosecution."
↩ - 21
Prosecutor v. Tadić, Case No. IT-94-1-AR72, Decision on the Defence Motion for Interlocutory Appeal on Jurisdiction (Int'l Crim. Trib. for the Former Yugoslavia Oct. 2, 1995). The Tadić Appeals Chamber's reasoning on jurisdictional legality and the customary status of international humanitarian law remains foundational for the doctrine's subsequent development.
↩ - 22
Prosecutor v. Akayesu, Case No. ICTR-96-4-T, Judgment (Int'l Crim. Trib. for Rwanda Sept. 2, 1998). The Akayesu judgment is the first international conviction for genocide and remains the seminal authority on the specific intent requirement (dolus specialis) and the use of rape as an act of genocide.
↩ - 23
Prosecutor v. Bemba Gombo, ICC-01/05-01/08, Judgment Pursuant to Article 74 of the Statute, Trial Chamber III (March 21, 2016). The Trial Chamber convicted Bemba under Article 28(a) for crimes committed by Mouvement de libération du Congo forces in the Central African Republic from October 2002 to March 2003.
↩ - 24
Prosecutor v. Bemba Gombo, ICC-01/05-01/08 A, Judgment on the Appeal of Mr. Jean-Pierre Bemba Gombo against Trial Chamber III's "Judgment Pursuant to Article 74 of the Statute," Appeals Chamber (June 8, 2018), 3–2 decision acquitting Bemba. For the dissenting view, see Dissenting Opinion of Judge Sanji Mmasenono Monageng and Judge Piotr Hofmański.
↩ - 25
For critical scholarly response, see Diane Marie Amann, "In Bemba and Beyond, Crimes Adjudged to Commit Themselves," EJIL: Talk! (June 13, 2018); Leila N. Sadat, "Prosecutor v. Jean-Pierre Bemba Gombo," 113 American Journal of International Law 353 (2019); Susana SáCouto & Patricia Sellers, "The Bemba Appeals Judgment: Impunity for Sexual and Gender-Based Crimes," 27 William & Mary Bill of Rights Journal 599 (2019).
↩ - 26
Genocide Convention, supra note 15, art. II(e); see also Prosecutor v. Krstić, Case No. IT-98-33-A, Judgment, Appeals Chamber, Int'l Crim. Trib. for the Former Yugoslavia (Apr. 19, 2004), upholding the conviction for aiding and abetting genocide based on the destruction of the Bosnian Muslim population of Srebrenica as a "substantial part" of the protected group within a specific geographic area.
↩ - 27
Office of the Prosecutor of Ukraine, statement quoted in Tom Dannenbaum, "Mechanisms for Criminal Prosecution of Russia's Aggression Against Ukraine," Just Security, March 10, 2022 (identifying child deportations as among "the most promising" avenues for proving genocidal intent in the Ukraine context).
↩ - 28
Rome Statute, art. 8 bis, as amended at Kampala in 2010 and activated July 17, 2018. The provision defines an act of aggression and limits jurisdiction over the crime to acts manifestly violating the UN Charter by their character, gravity, and scale.
↩ - 29
On the customary status of the crime of aggression and the structural limitations of Article 8 bis jurisdiction, see Carrie McDougall, The Crime of Aggression Under the Rome Statute of the International Criminal Court, 2nd ed. (Cambridge: Cambridge University Press, 2021); Tom Dannenbaum, The Crime of Aggression, Humanity, and the Soldier (Cambridge: Cambridge University Press, 2018).
↩ - 30
See Statute of the Special Tribunal for the Crime of Aggression against Ukraine, arts. 1, 6, 12–14, annexed to the Council of Europe–Ukraine Agreement of June 25, 2025. The Enlarged Partial Agreement was adopted by thirty-six states and the European Union at the 135th Session of the Committee of Ministers of the Council of Europe in Chișinău on May 15, 2026.
↩ - 31
See generally International Criminal Court, "Cases: Situation in Ukraine," https://www.icc-cpi.int/situations/ukraine. Subsequent warrants issued June 24, 2024, against Sergei Kuzhugetovich Shoigu and Valery Vasilyevich Gerasimov address alleged international crimes committed between at least October 10, 2022, and March 9, 2023.
↩ - 32
Rome Statute, art. 86 (general obligation to cooperate); art. 87 (requests for cooperation: general provisions); art. 89 (surrender of persons to the Court).
↩ - 33
President Vladimir Putin visited Mongolia, a State Party to the Rome Statute, on 2–3 September 2024 without arrest. On 24 October 2024, ICC Pre-Trial Chamber II found, in ICC-01/22, that Mongolia had failed to comply with the Court's request for cooperation pursuant to Article 87(7) of the Rome Statute and referred the matter to the Assembly of States Parties. See "Ukraine situation: ICC Pre-Trial Chamber II finds that Mongolia failed to cooperate in the arrest and surrender of Vladimir Vladimirovich Putin and refers the matter to the Assembly of States Parties," ICC Press Release, October 24, 2024.
↩ - 34
Yale Humanitarian Research Lab, Conflict Observatory program documents accessible at https://hub.conflictobservatory.org/. The Conflict Observatory operates with funding from the U.S. Department of State's Bureau of Conflict and Stabilization Operations and partnerships with Esri, the Smithsonian Cultural Rescue Initiative, and Planetscape Ai, among others.
↩ - 35
Nathaniel A. Raymond, executive director of the Yale HRL, has repeatedly emphasized in public testimony and academic writing that the Lab's findings have shaped the underlying evidentiary record relied upon in the ICC's Ukraine warrants. See also coverage in the Yale School of Public Health magazine, "Nowhere to Hide" (2023), describing the Lab's relationship to the Conflict Observatory.
↩ - 36
Reporting in 2025 documented the withholding of congressionally approved funds and the consequent risk to the Yale HRL's continued operation. See International Women's Media Foundation, "6,000 Pages of War Crimes: The Fate of Ukrainian Children" (2025), https://www.iwmf.org/reporting/6000-pages-of-war-crimes-the-fate-of-ukrainian-children/, citing Raymond's warning that "right now, it looks like our doors will close on July 1." For the underlying institutional account, see Testimony of Nathaniel A. Raymond, Executive Director, Humanitarian Research Lab, Yale School of Public Health, before the United States Senate Committee on Appropriations, available at https://www.appropriations.senate.gov/download/testimony-of-nathaniel-raymond. See also Ray Furlong, "Exclusive: 'A Catastrophic Blow' As US Shuts Unit Investigating War Crimes In Ukraine," Radio Free Europe/Radio Liberty, March 19, 2025.
↩ - 37
OSCE Permanent Council Decision invoking the Moscow Mechanism, March 3, 2022 (forty-five participating States); see OSCE/ODIHR, supra note 11.
↩ - 38
Council of Europe, Statute of the Register of Damage Caused by the Aggression of the Russian Federation against Ukraine, Committee of Ministers Resolution CM/Res(2023)3, May 17, 2023. The Register's Annual Activity Report for 2025 recorded 92,980 submitted claims and 30,461 recorded claims by the end of 2025. Council of Europe reporting on April 30, 2026 indicated that almost 150,000 claims had been submitted and more than 45,000 had been recorded.
↩ - 39
Convention Establishing an International Claims Commission for Ukraine, Council of Europe treaty, opened for signature 2025.
↩ - 40
See Christian Ritscher, "COVID-19 and International Crimes Trials in Germany," 18 Journal of International Criminal Justice 1077 (2020), discussing the procedural and institutional conditions under which Germany has conducted international crimes prosecutions. The German "structural investigations" model assembles evidence about the systematic conduct of a state or armed group as a foundation for subsequent individual prosecutions. Ritscher, formerly Head of the German War Crimes Unit S4 in the Office of the Federal Prosecutor General, was subsequently appointed Special Adviser and Head of the United Nations Investigative Team to Promote Accountability for Crimes Committed by Da'esh/ISIL (UNITAD) in September 2021.
↩ - 41
Higher Regional Court of Koblenz (Oberlandesgericht Koblenz), Judgment of January 13, 2022, in the case of Anwar Raslan (the so-called Al-Khatib trial). The first criminal verdict worldwide against a former Syrian state official for crimes against humanity, exercising universal jurisdiction under Section 1 of the German Code of Crimes against International Law (Völkerstrafgesetzbuch / CCAIL). The defendant was sentenced to life imprisonment for complicity in 27 murders, 4,000 cases of torture, and sexual violence as crimes against humanity. The German Federal Court of Justice rejected the appeal in August 2024, making the sentence final.
↩ - 42
Eurojust, Joint Investigation Team on Alleged Core International Crimes Committed in Ukraine, established March 25, 2022; expanded to include the International Centre for the Prosecution of the Crime of Aggression against Ukraine (ICPA) on July 3, 2023.
↩ - 43
United Nations Office of the High Commissioner for Human Rights, Berkeley Protocol on Digital Open Source Investigations (Geneva and Berkeley: OHCHR & Human Rights Center, UC Berkeley School of Law, 2022).
↩ - 44
Prosecutor v. Al-Werfalli, Case No. ICC-01/11-01/17, Warrant of Arrest (Aug. 15, 2017); Second Warrant of Arrest (July 4, 2018). On June 15, 2022, Pre-Trial Chamber I terminated proceedings following the Prosecution's notification of Al-Werfalli's death.
↩ - 45
On the role of Al-Werfalli in the development of digital evidence standards at the ICC, see generally Lindsay Freeman, "Digital Evidence and War Crimes Prosecutions: The Impact of Digital Technologies on International Criminal Investigations and Trials," Fordham International Law Journal (2018); Alexa Koenig, of the Human Rights Center at UC Berkeley School of Law, has written extensively on the verification challenges of social-media-derived evidence, including in work that contributed to the Berkeley Protocol on Digital Open Source Investigations.
↩ - 46
Rebecca J. Hamilton, "The Hidden Costs of Open-Source Evidence," Texas National Security Review (Fall 2023), discussing chain-of-custody, authentication, and the vulnerability of digital records to platform-side deletion.
↩ - 47
See generally Christoph Safferling, International Criminal Procedure (Oxford: Oxford University Press, 2012), on in absentia proceedings and fair trial standards; Rome Statute, art. 63(1) ("The accused shall be present during the trial").
↩ - 48
International Covenant on Civil and Political Rights, December 16, 1966, art. 14(3)(d), 999 U.N.T.S. 171 (right of the accused "[t]o be tried in his presence"); Sejdovic v. Italy, App. No. 56581/00, Eur. Ct. H.R. (Grand Chamber, March 1, 2006) (in absentia conviction must be followed by a right to retrial on the accused's presence).
↩ - 49
Prosecutor v. Al-Bashir, ICC-02/05-01/09-397, Judgment in the Jordan Referral re Al-Bashir Appeal, Appeals Chamber (May 6, 2019), ¶¶ 1, 113 ("There is neither State practice nor opinio juris that would support the existence of Head of State immunity under customary international law vis-à-vis an international court.").
↩ - 50
Cour de cassation (France), Assemblée plénière, Arrêt du 25 juillet 2025, in the case concerning the November 14, 2023 arrest warrant issued against Bashar al-Assad, then-President of Syria. The Court upheld the principle of absolute personal immunity (immunité ratione personae) for sitting foreign heads of state while indicating that Assad's loss of office allowed new warrants to be issued on a different immunity footing. The ruling stands in tension with the ICC Appeals Chamber's position in Al-Bashir.
↩ - 51
Statute of the Special Tribunal for the Crime of Aggression against Ukraine, arts. 12–14, addressing immunities and proceedings in the absence of the accused; see also Beth Van Schaack and Tom Dannenbaum, "Building a Special Tribunal for the Crime of Aggression against Ukraine," EJIL: Talk! (July 25, 2025).
↩ - 52
Compare International Criminal Tribunal for the former Yugoslavia, established by U.N. Security Council Resolution 827 (1993), operating until December 31, 2017; and International Criminal Tribunal for Rwanda, established by U.N.S.C. Resolution 955 (1994), operating until December 31, 2015 — both bodies' active jurisdictional life exceeded twenty years.
↩ - 53
On the politics of memory in post-conflict societies, see Jeffrey K. Olick, The Politics of Regret: On Collective Memory and Historical Responsibility (New York: Routledge, 2007); Mark Osiel, Mass Atrocity, Collective Memory, and the Law (New Brunswick: Transaction Publishers, 1997).
↩ - 54
The author's family was in Mariupol during the opening phase of the siege, living in the western part of the city in the district historically known as Zhovtnevyi (renamed Tsentralnyi in 2016 as part of decommunization, though Russian occupation forces have reverted to pre-2016 administrative names). The methodological position adopted here is that lived proximity to events does not exempt an author from analytical rigor; rather, it obliges the author to identify what proximity contributes and what it cannot adjudicate. Where the author's recollection corroborates institutionally documented facts — for example, the cutting of municipal services within the first days of the siege — this is noted in the body of the text. Where the author has no direct knowledge of an event, or where memory cannot be relied upon, the analysis draws exclusively on publicly available institutional sources.
↩ - 55
See, e.g., the Spanish Audiencia Nacional's exercise of universal jurisdiction under Ley Orgánica 6/1985, art. 23.4; Sweden's prosecutions under Chapter 22, § 6 of the Swedish Penal Code; and the Netherlands' International Crimes Act of 2003.
↩ - 56
On gravity, complementarity, and case selection at the ICC, see generally Margaret M. deGuzman, "Gravity and the Legitimacy of the International Criminal Court," Fordham International Law Journal (2009).
↩ - 57
On the theory and practice of universal jurisdiction, see generally M. Cherif Bassiouni, Introduction to International Criminal Law, 2nd rev. ed. (Leiden: Martinus Nijhoff, 2013); Antonio Cassese, International Criminal Law, 3rd ed. (Oxford: Oxford University Press, 2013).
↩ - 58
Bellingcat, Online Open Source Investigation Toolkit (updated annually). Bellingcat's methodology, including geolocation, chronolocation, and reverse-image verification, has set de facto standards in the open-source investigation community and has been integrated into multiple legal and quasi-legal proceedings.
↩ - 59
On the relationship between memory and reconstruction, see Jelena Subotić, Yellow Star, Red Star: Holocaust Remembrance after Communism (Ithaca: Cornell University Press, 2019); Lea David, The Past Can't Heal Us: The Dangers of Mandating Memory in the Name of Human Rights (Cambridge: Cambridge University Press, 2020).
↩ - 60
Lisa Parks, Cultures in Orbit: Satellites and the Televisual (Durham: Duke University Press, 2005), and "Vertical Mediation and the U.S. Drone War in the Horn of Africa," in Life in the Age of Drone Warfare, ed. Lisa Parks and Caren Kaplan (Durham: Duke University Press, 2017). Parks's broader argument is that the apparent transparency of overhead imagery obscures the chain of technical, institutional, and interpretive decisions that produce it; the legal implication, not addressed by Parks but relevant here, is that the evidentiary weight of satellite material in tribunal proceedings depends on a chain of trust the tribunal itself is poorly positioned to audit.
↩ - 61
On the anthropology of violence and the limits of documentary reconstruction, see Veena Das, Life and Words: Violence and the Descent into the Ordinary (Berkeley: University of California Press, 2007); on the related problem of how survivor testimony interacts with institutional record-making, see Das, "Language and Body: Transactions in the Construction of Pain," Daedalus 125, no. 1 (1996): 67–91.
↩ - 62
For the Ukrainian-language scholarly engagement with the conflict's legal aftermath, see Mykola Hnatovskyy, "Міжнародне гуманітарне право та збройний конфлікт в Україні" [International Humanitarian Law and the Armed Conflict in Ukraine], in Український щорічник міжнародного права [Ukrainian Yearbook of International Law] (Kyiv, 2023); see also publications of the Kharkiv Human Rights Protection Group on the documentation of civilian harm.
↩ - 63
Coalition for Content Provenance and Authenticity (C2PA), Technical Specification 1.x, content authenticity initiative documentation; see also Henry Farid, "Creating, Using, Misusing, and Detecting Deep Fakes," Journal of Online Trust and Safety 1, no. 4 (2022), on the technical landscape of media authentication.
↩ - 64
The author's recollection of the first days of the siege, particularly the collapse of cellular and electrical infrastructure in the western districts and the emergence of informal communication networks, is offered here under the methodological caveat described in Part I, supra note 54, and not as a substitute for documented sources.
↩ - 65
On the design of long-horizon institutional funding for accountability mechanisms, see the funding structures of the International, Impartial and Independent Mechanism on Syria (IIIM), established by UN General Assembly Resolution 71/248 (December 21, 2016), and funded through voluntary contributions pooled through the United Nations. On the Yale HRL funding interruption and subsequent donor-supported extension through October 1, 2025, see Yale School of Public Health, 'Donors Help Humanitarian Research Lab Keep Operating,' 2025.
↩ - 66
Kaveh Khoshnood and Nathaniel A. Raymond et al., "Damage Assessment of Health and Educational Facilities in Sievierodonetsk Raion, Ukraine: Evidence of Widespread, Indiscriminate, and Persistent Bombardment by Russia and Russia-Aligned Forces Between 24 February - 13 June 2022," Yale School of Public Health, Humanitarian Research Lab, issued 29 June 2022, Appendix I, Figure 1 (worked example of observable waves of damage; satellite imagery from 29 March 2022, 16 May 2022, and 30 May 2022) and accompanying methodological text ("It is not possible to conclude exactly how many separate instances of damage occurred to a facility given two images separated by an extended time period"). The Sievierodonetsk report is cited here for its methodological transparency, not because the Sievierodonetsk record substitutes for the Mariupol record; the same methodological constraints apply across Yale HRL's Ukraine portfolio.
↩ - 67
Kaveh Khoshnood, Nathaniel A. Raymond et al., "Evidence of Widespread and Systematic Bombardment of Ukrainian Healthcare Facilities: Summary," Humanitarian Research Lab, Yale School of Public Health, 17 May 2022. The published summary documents 22 of 277 facilities reviewed across five cities (Mariupol, Kharkiv, Kyiv Oblast, Chernihiv, and Izyum); of the 22 Mariupol facilities reviewed, the summary statistics table indicates 7 facilities with imagery-corroborated damage and open-source cross-corroboration. The full report was provided to the OSCE Moscow Mechanism and is available to international accountability mechanisms but not to the public.
↩ - 68
Humanitarian Research Lab at Yale School of Public Health, "System of Filtration: Mapping Russia's Detention Operations in Donetsk Oblast," Conflict Observatory Report, 25 August 2022. The figure of 1,700,000 transferred civilians, including 276,000 children, is attributed in the Yale report to the former Ukrainian Ombudswoman for Human Rights as of 14 June 2022. The Russian government figure of 3,400,000, including over 550,000 children, was reported on 16 August 2022. The methodological statement that "ground investigations are required to cross-corroborate these initial findings" appears in the Executive Summary of the Yale filtration report.
↩ - 69
Human Rights Watch and Truth Hounds, "Our City Was Gone: Russia's Devastation of Mariupol, Ukraine," with 3D reconstructions and visual and spatial analysis by SITU Research (New York: Human Rights Watch, 2024). The report draws on 240 interviews conducted between March 2022 and January 2024, analysis of dozens of high and very-high-resolution commercial satellite images, and over 850 verified photographs and videos. The 10,284-burial estimate is presented in Chapter VIII, derived from an analysis of grave sites at five cemeteries (Starokrymske, Novotroitske, Vynohradne, Pavlov Street, and Manhush) using individual graves visible on satellite imagery in combination with plaques and crosses on trench-style graves visible only in drone footage and on-the-ground photography. The report explicitly notes that this figure is likely a significant underestimate of total wartime mortality. The methodological integration of satellite-derived and testimonial layers, including the use of 3D photogrammetric reconstructions during witness interviews, is described in the report's Methodology section.
↩ - 70
Human Rights Watch and Truth Hounds, "Our City Was Gone," Chapter X (Russian Chain of Command). The seventeen identified units include elements of the 8th Combined Arms Army, the 150th Motorized Rifle Division and its 68th Tank Regiment, the 102nd Motorized Rifle Regiment, the 810th Naval Infantry Brigade and the 382nd Naval Infantry Battalion subordinated to the Black Sea Fleet, the 22nd and 346th Separate Special Purpose Brigades, the 141st Special Motorized Regiment and the Special Purpose Police Regiment from Chechnya, and multiple units from the so-called Donetsk People's Republic 1st Army Corps. The ten named individuals identified as potentially bearing command responsibility are Vladimir Putin, Sergei Shoigu, Valery Gerasimov, Sergei Rudskoy, Alexander Dvornikov, Viktor Zolotov, Andrei Mordvichev, Ramzan Kadyrov, Adam Delimkhanov, and Denis Pushilin. The methodology for the chain-of-command analysis is set out in the same chapter.
↩