A Single Comment — Permalink
© ICCforum.com, 2010–2023. All rights reserved. Policies | Guidelines
- AhmadSoliman: Admissibility of Illegally Obtained Cyber Evidence at the ICC I. Introduction The International Criminal Court (ICC) is already grappling with questions about what types of cyber evidence will be admissible to the Court. Even more difficult questions about the admissibility of illegally obtained cyber information are on the horizon. The law governing the admissibility of illegally obtained cyber evidence contains several ambiguities, and new types of... (more)
- Steve Helmeci: The “Coming Storm”: Possibilities for Preserving Evidence of War Crimes Published on Major Online Service Providers I. Introduction There should be no greater windfall for criminal prosecutors, national or international, than the growth of the Internet as a tool for communication and networking. Potential evidence against perpetrators abounds: photos and videos posted by offenders, victims, and witnesses alike; personal... (more)
- glazera2020: Protecting Against Deepfakes: How the ICC Can Ensure Trust in the Verification and Use of Open-Source Evidence I. Introduction Digital devices have allowed ordinary civilians to become on-the-ground investigative reporters in almost every region of the world. Contemporaneous uploads of footage featuring protestors being tear-gassed in Hong Kong or Syrians fleeing from chemical warfare have allowed viewers all over the... (more)
- brittnewell: What Policies Should the Office of the Prosecutor Adopt in Receiving Cyber Evidence From User-Generated Evidence Gathering Apps to Help Protect Those Providing It? I. Introduction The rapid increase in the use of camera-equipped and internet-connected devices has enabled individuals to record far more information about their lives and their surroundings than ever before. This fundamental shift in information gathering is transforming... (more)
- Abhishek: Digital Evidence and the Use of Artificial Intelligence I. Introduction With the advent of new advancements in the technological world, it should come as no surprise that the primary source of information gathering and documentation is within the grasp of every individual. Internet and smartphones have virtually created journalists and investigators on every corner. International human rights and criminal prosecution organizations have... (more)
- marianava: Digital Evidence Repositories and Vulnerable Populations: How the Accumulation of Digital Evidence May Interact with the Privacy of Sexual Assault Survivors I. Introduction With the rise of digital technology, it is likely that the future of evidence collection at the International Criminal Court (ICC) will increasingly rely on digital evidence. This may take the form of videos, images, etc. (media) taken by victims or civil society... (more)
- Jill Mierke: The Promise and Problems of Open Source Evidence in ICC Investigations and Trials I. Introduction On August 15, 2017, the Pre-Trial Chamber of the International Criminal Court (ICC) issued an arrest warrant for senior Libyan military commander, Mahmoud Al-Werfalli.1 Al-Werfalli is accused of personally committing or ordering thirty-three murders in Benghazi between June 2016 and July 2017. Importantly, this is the... (more)
Comment on the Cyber Evidence Question: “To what extent can cyber evidence repositories, and digital and open-source evidence, facilitate the work of the OTP, and the ICC more generally?”
Digital Evidence and the Use of Artificial Intelligence
With the advent of new advancements in the technological world, it should come as no surprise that the primary source of information gathering and documentation is within the grasp of every individual. Internet and smartphones have virtually created journalists and investigators on every corner. International human rights and criminal prosecution organizations have had their eyes for quite some time on the potential benefits of using the open web and social media platforms to gather information and evidence for investigation, prosecution, and averment of crimes. It has already been seen that International Tribunals such as the International Criminal Tribunal for Rwanda (ICTR), the International Criminal Tribunal for former Yugoslavia (ICTY) and even the International Criminal Court (ICC) have initiated investigations and prosecuted accused persons on the basis of digital evidence. With the benefits of having such vast amounts of information at hand there are bound to be challenges as well. The key challenges in using digital evidence are with regards to identification of relevant evidence and establishing their authentication in order to satisfy the Court’s threshold of admittance.
As we have seen in the case of Sri Lanka’s soldier’s 2009 video capturing execution of prisoners, forensic analysts have been used to verify the authenticity of digital evidence,1 but there has been very limited success in identifying the correct evidence. The relevance of evidence has been established only through manual sorting and this potentially leaves a lot of ground to be covered. Artificial Intelligence (AI) can be employed in such scenarios to effectively scour a dump of data and comb out the relevant information. In this comment, I first briefly discuss the threshold for admissibility, and standards for assigning evidentiary value to digital evidence. Then I discuss the potential of AI in gathering and sorting data to be used for investigation and prosecution by the ICC.
II. Admissibility of Digital Evidence
The ICTR, the ICTY or the ICC do not have a separate criterion for regulating digital evidence and it comes under the broad definition of documentary evidence.2 The ICC derives its admissibility criteria for evidence from Article 69 of the Rome Statute which requires that the Court takes the probative value of the evidence into account while deciding on its admissibility. At this stage, the evidence needs to only prove its prima facie credibility. However, the weight of the evidence gets decided at the conclusion of trial in light of all the evidence.3
These tribunals have generally relied upon supporting eyewitness testimony to establish the credibility of digital evidence. In the Milan Milutinović trial, the chamber relied on the testimony of camera operators to establish the reliability of footage provided by the BBC.4 However, this has not stopped the tribunals from allowing admission of evidence even if it is not supported by witness testimony. The ICTY has held that there is no blanket ban on the admission of documentary evidence simply due to the lack of supporting witness testimony.5 A similar stand was taken by the ICC when it ruled:
Even though the threshold for evidence at the admissibility stage is quite low, the ICC still has certain guidelines for digital evidence. It has been ruled that digital evidence must conform to “e-Court Protocol”7 to “ensure authenticity, accuracy, confidentiality and preservation of the record of proceedings.”8
III. Evidence Gathering
The question of the admissibility or evidentiary value arises only after the relevant evidence has been collected. The collection of evidence proves to be the most challenging part as these cases revolve around gross violation of human rights and it sometimes takes years even to start the investigation. In such situations, it becomes imperative that any form of evidence be collected and preserved as soon as it becomes available. Many initiatives, by NGOs and private organizations, have been started in order to collect, investigate, and preserve digital evidence such as video files or images. Some of these are:
BellingCat: They are an independent team of investigators, journalists and researchers who use open source information such as images or video on social media sites to investigate conflict or human rights violation. They have already worked on the chemical attack in Syria and the downing of the MH17 flight.9
Syrian Archive: They too gather information from open source social media platforms and other websites and preserve the relevant data for further use.10
WITNESS: They support and train people in conflict zones to “capture and preserve footage to improve the chances of it being used in courtroom.”11 They also operate the WITNESS Media Lab which works towards verifying and authenticating the videos.
ACLU Mobile Justice App: The American Civil Liberties Union has launched the Mobile Justice app which can be used to record and upload any unwarranted interactions with law enforcement.12 These videos are then anonymously sent to the ACLU staff.
These initiatives are definitely a step in the right direction but, given the fact they are still very limited in scope and availability, have not been able to fulfill the ICC’s requirements for credible sources of digital evidence. These can only go as long as to provide evidence as amicus curiae or in the capacity of a friend to the court, and their evidentiary value would be decided in light of all other pieces of evidence available with the Court. If the ICC is to truly become a pioneer in using digital evidence to investigate and prosecute crimes at the international level, then there is a need to form its own archive or repository where such websites or individuals can submit any form of videos or images against human rights violations. There would certainly be challenges to such an initiative, but I believe benefits will outweigh these challenges. The challenges faced by the ICC would include satisfying non-violation human rights as per Article 69(7) of the Rome Statute, privacy concerns, safekeeping of the materials, establishing authenticity, objectively weeding out crucial evidence, etc. In this comment, I only discuss meeting the challenge to weed out important pieces of evidence amongst an overwhelming amount of data and how AI can be employed for such purposes.
IV. Artificial Intelligence
Almost all the initiatives I mentioned before, and many other as well, manually sort out the relevant piece of information amongst the data that is uploaded or collected by their respective organizations. This is a very tedious process and requires a lot of man hours. Additionally, there is always the risk of leaving out many crucial pieces of evidence. To put things into perspective, consider the example of the Syrian Archive which had 3,314,265 uploads on December 15th, 2019 and only 5,743 of these were verified.13 Now one can only imagine the number of videos or images that would be uploaded if the ICC decided to create its own repository. In such a scenario, I believe the first order of business would be to objectively weed out the relevant pieces of information and only then initiate work on establishing authenticity, or chain of custody, etc. This is where AI comes into play.
AI can be understood as a software or tool designed:
Another major application of AI comes in the form of Machine Learning which tries to replicate the human ability to learn from experience.
It is this ability of Machine Learning which makes it particularly useful for the criminal justice system. The potential uses of AI in order for the ICC to conduct an effective investigation into the crimes under its purview are as follows:
A. Weeding Out Important Evidence
If the ICC is to create its own repository or even use the existing repositories created by several other organizations then it becomes very hard for them to weed out the relevant and crucial piece of evidence. However, this is exactly the scenario where AI and Machine Learning shines. Data has often been described as the lifeblood of AI and the more data the better it is for AI. Many NGOs are already using AI to scour and report child pornography.15 The AI used by these organizations have algorithms designed to recognize features or parameters similar to those found in child pornography.16 Similarly, an algorithm can be developed to identify different patterns of violence or human rights violations. These will not only save countless human hours but will also enhance productivity.
B. Gathering Information from Different Platforms by Recognizing Patterns
AI can also be used even if the ICC does not create its own repository or partner with any other organizations maintaining such repositories. A similar AI as described above can be used to crawl through those websites which allow crawling or scouring and can collect potential pieces of evidence, identifying them on the basis of parameters set by the developers.
C. Detection or Prediction
Another use of AI is the detection or prediction of incidents which violate international law. This might not always come in the general scope of the ICC but can still be useful for gathering evidence. An example of such use is the “Sentry” system which provides Syrian civilians early warning of incoming airstrikes.17 Its developer, Hala Systems, describes Sentry as an “indication and warning system that utilizes a multi-sensor network to generate a credible, real-time, situational awareness of threats.”18 It furthers explains that “Sentry uses AI to instantaneously validate information from multiple sources, allowing stakeholders to detect, identify, and predict threats.”19
V. Challenges of Employing AI
Although the term “Artificial Intelligence” was coined in 1956 by John McCarthy,20 it was not until 2010 that real progress on AI started.21 Being such a new and revolutionary technology comes with its own set of challenges that will have to be addressed in order to reap in maximum benefits of this technology. A few of the key challenges are discussed below.
A. Accuracy and Error
As mentioned earlier, AI uses certain parameters which has been set by the developer to identify the required patterns. Consider, for example, how facial recognition works. Facial recognition “uses biometrics to map facial features.”22 If these features match the features present in its database, then a match can be found. Even if the tool is not looking for match, it will still use certain features to establish whether the input is a human face or not. The problem arises in cases where the image or video might not be clear enough, or the feature of the face might not be consistent with the parameters inputted. In such scenarios, there would be an error in the results.
This would be even more challenging in case of the ICC because it is very possible that the people capturing videos and images in conflict zones might not have access to capable equipment or even a very good camera phone and this is bound to result in very low quality and even distorted videos or images.
In order to counter this challenge, there are two options which can be employed by the ICC. First and foremost is the obvious improvement in the AI’s capacity given its machine learning abilities. Accordingly, the more data that is provided to the AI, the better the results it will produce. The second option would be to follow the works of organizations like Witness who have been training and advising ordinary citizens on how to capture better videos and images of conflicts.
In the event that the ICC employs AI to scour or crawl social media websites to gather potential evidence, it is certain that it would need access to major social media sites such as Facebook, YouTube, Twitter, etc. But the issue here arises from the fact that majority of these websites are US headquartered and follow US rules. Now, given the fact that the US administration has openly criticized the ICC and has made it clear that they will not support ICC in any way, and neither will allow its companies to support the ICC,23 makes it impossible for the ICC to get access to these sites. This might be overcome if these companies’ entities are established in countries outside US and they allow access to the ICC but given the administration’s threat of sanctions it is very improbable that the companies will go forward with such schemes.
C. Lack of Resources and Expertise
AI is very sophisticated technology and it has not developed enough to be used by untrained individuals. This basically translates into the requirement of a lot of expertise and resources in order to develop and operate any piece of AI. Given the fact that the ICC’s budget and annual expenditure is constantly being questioned by the member countries,24 it would be quite challenging to undertake a project to develop AI for its own use. This issue could be resolved if the ICC partners with one of the tech companies already possessing AI. Such partnerships have proved very fruitful for many NGOs who received free access to AI tools to curb child sexual content.
The ICC has been facing many issues with regard to the use of digital evidence and AI is the way by which it can enhance its capability to gather and authenticate evidence for investigation and prosecution. The potential of AI can be understood from the fact that a ton of research and studies are already being conducted on the uses of AI in furtherance of Criminal Justice needs and it is time the ICC too joined the bandwagon. Having said that, I do accept that there are many challenges to overcome and it will take lot of time to iron out all the wrinkles. In conclusion, I would like to point out the fact that the trends of criminal investigation are evolving at a breakneck speed and, in order to stay relevant, the ICC will have to evolve too.
Endnotes — (click the footnote reference number, or ↩ symbol, to return to location in text).
Louis Charbonneau, Sri Lanka Execution Video Appears Authentic—U.N., Reuters, Jan. 7, 2010, available online. ↩
The Prosecutor v. Édouard Karemera, Mathiey Ngirumpatse and Joseph Nzirorera, ICTR-98-44-T, Decision on the Prosecutor’s Motion for Admission of Certain Exhibits Into Evidence (ICTR TC III, Jan. 25, 2008), available online. ↩
The Prosecutor v. Jean-Pierre Bemba Gombo, ICC-01/05-01/08, Decision on the admission into evidence of materials contained in the prosecution’s list of evidence (TC III, Nov. 19, 2010), available online. ↩
The Prosecutor v. Milan Milutinović, Nikola Šainović, Dragoljub Ojdanić, Nebojša Pavković, Vladimir Lazaarević and Serten Lukić, IT-05-87-T, Judgement, ¶ 896 (ICTR TC III, Feb. 26, 2009), available online. ↩
The Prosecutor v. Radoslav Brđanin and Momir Talić, IT-99-36-T, Order on the Standards Governing the Admission of Evidence, ¶ 20 (ICTY TC II, Feb. 15, 2002), available online. ↩
The Prosecutor v. Thomas Lubanga Dyilo, ICC-01/04-01/06, Decision on the confirmation of charges, ¶ 96 (PTC I, Jan. 29, 2007), available online. ↩
The Prosecutor v. Callixte Mbarushimana, ICC-01/04-01/10, Decision amending the e-Court Protocol (PTC I, Apr. 28, 2011), available online. ↩
About, Bellingcat, available online (last visited May 29, 2020). ↩
About, Syrian Archive, available online (last visited May 29, 2020). ↩
Our Work, WITNESS, available online (last visited May 29, 2020). ↩
ACLU App to Record Police Conduct, ACLU, available online (last visited May 29, 2020). ↩
Observation Database, Syrian Archive, available online (last visited May 29, 2020)
(providing no information on the number of videos that have actually been checked). ↩
See Matthijs M. Maas, International Law Does Not Compute: Artificial Intelligence and the Development, Displacement or Destruction of the Global Order, 20 Melb. J. Int’l L. 29 (Aug. 2019), available online. ↩
Angelin Yeoh, Google gives NGOs the Power of AI to Detect Child Sexual Abuse Images, The Star, Sep. 4, 2018, available online. ↩
Danny Gold, Saving Lives with Tech Amid Syria’s Endless Civil War, Wired, Aug. 16, 2018, available online. ↩
Home Page, Hala Systems, available online (last visited May 29, 2020). ↩
Christopher Rigano, Using Artificial Intelligence to Address Criminal Justice Needs, 280 NIJ Journal 1 (Jan. 2019), available online. ↩
Steve Symanovich, How does Facial Recognition Work?, Norton, available online (last visited May 29, 2020). ↩
Owen Bowcott, Oliver Holmes & Erin Durkin, John Bolton Threatens War Crimes Court with Sanctions in Virulent Attack, The Guardian, Sep. 10, 2018, available online. ↩
Sophie van Leeuwen, Many ICC Countries are not Paying, Just. Hub (Jun. 22, 2015), available online. ↩