In the great majority of cases, state or non-state armed forces engaging in violations of international humanitarian law (IHL) do not attempt to justify their acts by offering different interpretations of the applicable law. Rather, what they usually do is simply flat-out deny that the facts alleged actually happened. When civilians get killed in the course of hostilities, for example, the actors concerned will only rarely provide a legal argument about (say) the notion of direct participation in hostilities, which can result in the removal of protection from attack. What they will do instead is argue that the attack never happened, or that it was done by some other party, or that it may have resulted from an honest mistake and that the civilians or civilian objects were not targeted intentionally.
Thus, most disputes involved alleged violations of IHL are disputes about the facts, not about the law. Actors engaged in conflict – even those that may, in principle, be committed to compliance with IHL – have a structural interest to generate alternate realities or lie, to local or global audiences alike. Accurate, reliable fact-finding is thus a crucial element (one among many) for incentivizing compliance with IHL and ensuring accountability.
Today, a wide array of actors, with different functions and purposes, attempt to establish facts relevant to alleged violations of IHL in real time, as the conflict unfolds. This includes the media (traditional or novel), civil society organizations, commissions of inquiry mandated by international organizations, military authorities, and national and international prosecutors. Such fact-finding efforts are inevitably diffuse, imperfect and overlapping. Equally inevitably, in highly polarized contexts – as most armed conflicts are – the factual findings of all of these actors may be contested and pushed back against, often through concerted propaganda.
In this post, we will critically examine a proposal for a new kind of mechanism, which would complement existing fact-finding efforts. The core idea of the proposal is to harness the power of modern computing technologies by establishing a global database of incidents implicating possible violations of IHL. This would be done by creating a platform through which various actors in the field, operating at different levels (e.g. trained professionals v. ordinary citizens), could submit evidence for inclusion into the database, including video, audio or documentary material. These submissions would then be filtered and processed by a bespoke AI system to check the reliability and nature of the information provided. Trained human reviewers would then verify the information collated by the AI system and approve the inclusion of specific incidents into the global database, which would be publicly available. We will first provide some background on existing fact-finding mechanisms and then explain what the new proposal could entail, and what its advantages and limitations would be.
Comparing human rights and IHL fact-finding
The importance of objective fact-finding has long been recognized for ensuring compliance with IHL. Indeed, it is recognized in the IHL treaties themselves. The four 1949 Geneva Conventions currently in force, like their immediate predecessors of 1929, provide for the use of an enquiry procedure into any alleged violations of the Conventions (see Art. 1 52 GC I, Art. 53 GC II, Art. 132 GC III, Art. 149 GC IV). But this enquiry procedure has never been successfully used, because it requires the consent of both parties to the international armed conflict.
The same applies to the institutionalized enquiry mechanism created by Article 90 of Additional Protocol I to the Geneva Conventions – the International Humanitarian Fact-Finding Commission (IHFFC). Article 90 was designed to provide an independent body capable of investigating alleged violations of the laws of war, offering a crucial mechanism for accountability during armed conflicts. However, despite its adoption into the text of the Additional Protocol, an enquiry procedure under Article 90 has never obtained consent from both parties to the conflict, due to the obvious lack of political will.
In recent years, the IHFFC has twice been invited to exercise ‘good offices’, rather than an enquiry under Article 90: first in 2017 regarding an incident in Eastern Ukraine, when it was invited by the OSCE to do a report, and secondly last year, when Poland invited the Commission to prepare a report on the 1 April 2024 Israeli strike on World Central Kitchen personnel in Gaza. In July, the Commission announced that its confidential report to the government of Poland was submitted back in October, but it is up to Poland to decide on whether the report will be published and it has not yet done so. (For more discussions of these efforts by the IHFFC, see Garraway here and Poulopoulou here).
Despite these two recent activations of the IHFFC on a good offices basis, it seems to us beyond dispute that Article 90 AP I has failed to live up to its promise as a tool for enhancing accountability and compliance with IHL – just like the enquiry procedure under the GCs. And while of course state-authorized international criminal tribunals have an essential role to play in reliably establishing facts relevant to IHL violations in war crimes prosecutions, which has been much more successful than that of the IHFFC, it is only one oriented on post facto accountability, often many years after a conflict is over, and not on real-time monitoring.
Resistance from states to more robust IHL-specific compliance mechanisms
The weakness of IHL-specific fact-finding mechanisms is a direct result of the lack of political will among states. We can observe the same type of resistance regarding softer, more general compliance mechanisms proposed a decade ago – efforts that were ultimately abandoned, reflecting deep political reluctance to institutionalize any form of collective scrutiny.
It is well known that the ICRC and Switzerland launched an initiative following Resolution 1 adopted at the 31st International Conference in 2011, which called for exploring ways to strengthen compliance monitoring. These attempts were grounded in the desire to ensure compliance with international humanitarian law and to mitigate civilian suffering by increasing transparency and accountability in wartime actions. Between 2012 and 2015, the initiative involved an extensive consultation process, including nine meetings with over 140 States, alongside bilateral and regional discussions. Each meeting was informed by careful research and background materials that refined options based on ongoing dialogue. Civil society and national Red Cross and Red Crescent Societies were also engaged to ensure diverse perspectives.
This inclusive and methodical process led to a Concluding Report and a draft resolution presented at the 32nd International Conference in 2015. However, a consensus could not be reached on the proposed mechanisms. Key issues included apprehensions about state sovereignty, the perceived politicization of the proposed mechanisms, and doubts about their effectiveness and enforceability. As a result, the initiative did not receive the necessary support to be adopted. Between 2015 and 2019, Switzerland and the ICRC jointly conducted an intergovernmental process, focusing on improving compliance with IHL. The main objective was to reach consensus on establishing a forum for dialogue among States on IHL and on ways to make better use of the International Conference and regional forums. However, despite extensive consultations, the process did not achieve consensus. (See more here and here).
In short, state-driven and IHL-specific fact-finding mechanisms are simply infeasible – and the political climate in the near-to-medium term does not look any more favourable. And this is a gap that must be filled. Armed conflicts are present in almost all continents of the world (with threats to expand soon into new areas such as the Artic), and there is a shocking and careless disregard for the rules of IHL even by actors who used to largely apply it.
Human rights fact-finding as a substitute?
It has long been recognized that fact finding efforts in the field of human rights have been fundamental for ensuring compliance at a global level. Grass-roots organizations and local NGOs, umbrella NGOs and international advocacy groups, regional human rights commissions and courts, U.N. experts and treaty-based committees, and a number of UN fact-finding missions compose a robust network of organizations gathering and using information on the day-to-day situation of human rights in the field, with the aim of enhancing and improving compliance in that field of international law. Although this certainly does not solve issues of non-compliance with human rights obligations, it enables the movement of information from the most remote, corrupt and often forgotten places, to the eyes of students, academics, activists, policy makers and diplomats all over the world.
Some of these human rights institutions have also taken up fact-finding relevant for IHL violations, and have applied IHL directly. But this monitoring landscape is inevitably imperfect and fragmentary – and using it as a model for developing IHL-specific mechanisms is not realistic.
Many human rights monitoring and documenting efforts refer to armed conflicts with high notoriety and intensity over long periods of time. For instance, the UN’s fact-finding mandates that include some IHL monitoring, which require an expenditure of political capital for their establishment and maintenance, only cover certain countries and conflicts. The service they provide, while invaluable, is simply scarce in comparison to the number of armed conflicts that exist in the world (with the ICRC counting ca. 120 armed conflicts in 2024). It should be highlighted that, due to the low threshold for the existence of an international armed conflict (IAC) there are many IACs that are ´short-lived´ and do not receive much attention and coverage from monitoring efforts (even at the domestic level). Border skirmishes between India and China or Cambodia and Thailand exemplify this. Similarly, there are many non-international armed conflicts (NIACs) that enter periods of lower intensity or are simply forgotten by monitoring efforts that also lead to the same result. The conflict in West Papua is a good example of this.
Another interesting facet of the same problem appears when looking at monitoring from a thematic point of view. Some organizations have shown very important efforts in trying to achieve a broad coverage, but usually this is possible over narrow areas of IHL. The work of the Aid Worker Security Database (AWSD) on attacks on humanitarian personnel is a good example of this. The approach is interesting as it paints a more detailed picture regarding a very real and important aspect highly relevant for IHL compliance. However, the number of initiatives doing this type of work is minimal when compared with the number of areas that IHL has. Evidently, not every conflict will show incidents regarding all types of IHL rules, but it is undeniable that some areas of IHL such as the conduct of hostilities, or the treatment of persons detained, receive more focus than less known provisions such as the proper treatment and disposal of the dead. It is necessary to incentivise further monitoring towards these areas.
Moreover, the mandates and methodologies of UN fact-finding missions are usually framed within the broader human rights framework rather than IHL. They focus on documenting patterns of violations such as extrajudicial killings, arbitrary detention, or restrictions on freedoms, which fall squarely under human rights law, and only secondarily address IHL breaches that arise in armed conflict. While these missions sometimes reference IHL when relevant—such as in cases of indiscriminate attacks or targeting of civilians—their primary orientation remains toward human rights norms and accountability. This emphasis can leave gaps in the systematic identification and legal analysis of conduct that constitutes IHL violations, particularly in areas that require high levels of specialization, underscoring the need for monitoring mechanisms that are specifically designed to assess compliance with the laws of armed conflict. And it is often difficult for these missions to recruit members and staff who have deep specific expertise in IHL.
IHL fact-finding: the way forward
Several recent initiatives complement one another in their shared aim of strengthening compliance with IHL worldwide. The ICRC’s Global Initiative seeks to promote a renewed political and legal dialogue on IHL compliance at the international level, fostering state engagement and practical commitments. The Geneva Academy’s IHL in Focus tries to provide legal analysis of contemporary IHL issues. The Beyond Compliance Consortium will bring together diverse stakeholders to generate innovative, multidisciplinary approaches that go beyond traditional legal mechanisms. The IHL Centre launched a practitioners’ guide to help field actors to conduct IHL field assessments and advocacy efforts. Finally, the Raoul Wallenberg Institute’s International Compliance Mechanism Database (ICMD) will offer a systematic mapping aided by AI for the analysis of trends and patterns regarding IHL in a data-driven way. Taken together, these initiatives reinforce each other and seek to advance the goal of more effective IHL compliance.
These efforts complete data-collection projects that are not specifically focused on IHL violations. Projects like ACLED (Armed Conflict Location & Event Data) and the Uppsala Conflict Data Program play a vital role in tracking patterns of political violence, conflict events, and civilian casualties, providing valuable datasets for understanding global conflict dynamics. However, their methodologies are not designed to focus on IHL, and as a result these projects measure the broader occurrence of violent incidents rather than whether such acts amount to potential IHL violations. Similarly, investigative journalism initiatives like Bellingcat, The New Humanitarian´s Aid Seekers Database, or BBC Verify contribute crucial evidence on conflict-related events through open-source verification, but their focus is on documenting and corroborating violent acts rather than systematically applying legal frameworks. Together, these efforts offer indispensable information, but they do not substitute for monitoring mechanisms tailored to identifying, classifying, and analyzing violations of IHL.
But there is still a need for a data-driven approach to documenting IHL violations worldwide. This is the ambition of the Raoul Wallenberg Institute´s ICMD project. Its aim is to use new technologies to gather, vet and flag data relevant to IHL in an efficient way that allows for big data analysis to look into trends and patterns at global and conflict levels. ICMD seeks to address some of the problems that have been highlighted throughout this post. For instance, ICMD seeks to create a network of global contributors and the infrastructure needed for discreet and easy access that would enable reporting from anywhere in the world (similar to the ICC´s Project Harmony or the IBA Eyewitness projects) to centralize fractioned data, particularly data coming from conflicts not heavily covered by media outlets.
The use of AI in this infrastructure would streamline the processing of information to achieve a fast system that keeps track in an efficient matter, instead of relying on old information. The selection, vetting, and flagging of information would be done within an IHL framework from the outset to tailor the information to a humanitarian audience.
For non-judicial fact-finding mechanisms, the relevant standard of proof is necessarily considerably lower than for proving individual criminal responsibility or state responsibility in a court of law. This is important to highlight. The aim of ICMD is to create a record of facts relevant for IHL that are flagged for subsequent engagement. It is not to suggest that the data entries are fully determined violations of IHL proven beyond any doubt. It would be for other actors to engage in formal proof, to whatever standard they need to apply for their own purpose. ICMD does not intend to fall prey to the dynamic of playing judge from an academic setting, only to be shut down whenever a party to a conflict suggests that its action was lawfully taken, based on confidential information that only it has available and will not release.
In other words, by keeping track of incidents that seem relevant for IHL, duty bearers might feel pressure (from authorities, other states, or civil society) to explain convincingly why they have, for example, attacked certain hospitals, ambulances or cultural sites. Most importantly, the process would be data-oriented with the sole aim of creating a record that provides up-to-date information, which can facilitate better policy-making, compliance and accountability efforts, and even education and training. This approach will provide very important data for other initiatives. The use of user-generated input, processed with the assistance of AI and creating a living, real-time picture of IHL worldwide would be worth the effort needed to build and maintain something that already exists in other areas such as democracy, arms trades and corruption.
Limitations
ICMD could have substantial added value when compared to existing fact-finding approaches – which it would of course complement, rather than try to supplant. In particular, it would make it much easier to identify patterns of conduct, within a certain conflict or globally. It would also facilitate the work of other fact-finders.
However, the success of the database depends greatly on the network of people and organizations on the ground willing to partake in this exercise. This, of course, requires efforts in engaging with civil society during the process of creating a safe and discreet mechanisms for contributing data in a way that does not pose a risk for anyone or restricts access to any organization. Similarly, efforts are required to maintain the data safe in an era of cyber-attacks, as well as preventing access from undesirable actors (including local authorities who might want to initiate fishing expeditions using the data).
Obviously, concerns may also arise as to the reliability of the AI and other technologies used – but such concerns are, in our view, mitigated by the fact that AI systems are already being used effectively at the ICC and elsewhere, and that expert human review would ultimately control all content. And, of course, the practical utility of this new fact-finding system can ultimately only be assessed once it is operational. Finally, political acceptance by stakeholders remains uncertain in the current political climate. However, we believe that a data-centred approach that distances itself from the goals of accountability and advocacy and focuses on trends and patterns can become useful for all interested parties, including governments, militaries and organizations planning operations on the ground, and a more general audience such as academics, students, journalists, and others interested in this topic.