Digital Accountability Symposium: Harnessing User-Generated Content in Accountability Efforts for International Law Violations in Yemen

[Dearbhla Minogue is a Legal Officer for the Global Legal Action Network and Ruwadzano Patience Makumbe is a Zimbabwean lawyer and currently a Hillary Rodham Clinton Global Challenges Programme Scholar at Swansea University.]

Digital technologies have empowered citizens to not only be consumers of information but also creators and distributors, making it possible for ordinary people to highlight human rights violations across the world. As has been the case in Syria, user-generated content (UGC) has played a significant role in exposing atrocities in Yemen. In the spirit of harnessing open-source intelligence (OSINT) as a source of evidence of violations of international law, Global Legal Action Network (GLAN) and it partners are examining the extent to which this information can be used in accountability processes.

OSINT will never replace witness testimony and field investigations, which are considered more reliable and can themselves provide crucial audiovisual evidence for prosecutions. However, it can supplement and corroborate such evidence. Take, for example, the attack on Zabid market in Hodeidah, Yemen, on 12 May 2015. Witnesses told Yemeni researchers that three large bombs landed in a crowded market, killing dozens of civilians. One survivor told field officers from Mwatana:  

“My brother and I were working in a shop selling bread when we heard the first and second explosions. We came out of the shop to see what happened, and people were scared, fleeing in all directions […w]hereas my brother rushed to rescue the injured in the attack. I was afraid that he may get hurt, and indeed the third bomb landed in the road but it did not explode, but fragments hit my brother in the head and neck when he was on his way to rescue people and he died on the spot. People had called on him not to go as he may get hurt, but he went to his death.”

In response to the allegations, the Saudi/UAE-led Coalition’s accountability body claimed that the Coalition was not responsible for this attack. The UK government and even the English courts have shown a huge degree of deference to such official sources – and this is where OSINT comes into play. Aftermath videos found by online investigators revealed widespread death and destruction and characteristic ‘pancaking’ – that is, the stacking of multiple floors on top of each other after support struts are blown out from the inside – of the Zabid market building. Another video revealed two distinct columns of smoke, which helped experts to conclude that two large air-delivered bombs were likely responsible for this damage – a significant finding given that the Coalition controls the airspace over this area.

Such information may create political discomfort for powerful actors such as Saudi Arabia who are unused to having their narratives challenged, but is it evidence? And what value does it add given that human rights NGOs have already investigated such incidents?

In 2017, an English Divisional Court, dismissing a challenge brought by Campaign Against the Arms Trade (CAAT) against the licensing of weapons sales to Saudi Arabia by the UK government for use in the war in Yemen, stated that ‘open source material [i.e. NGO reporting] is only part of the picture’ (para 86). It also found that the sources underpinning a United Nations report on Yemen, which relied extensively on interviews, were ‘not qualitatively as sophisticated as the sources available to the MoD’ (para 208(5)(e)). On 20 June 2019 the Court of Appeal set aside the Divisional Court’s order, accepting CAAT’s argument that the government’s decision-making procedure was flawed. While it also appeared to rightly row back on the lower court’s criticisms of NGO evidence (see para 134 in particular), the Court would likely afford a similar degree of deference (a standard feature of judicial review, particularly in cases of this nature) to official sources in any future challenge. Of course, the major obstacle will always be the inability of claimants to access classified information – in the Court of Appeal’s language, the “why” to accompany the “what”. But this is more reason for the accountability community to be ready to broaden the spectrum of evidence that is available for use in our efforts. Where OSINT material can prove something that is missing from a case hypothesis or can corroborate a witness’s testimony, it can help to build a stronger case and there is every reason to strive to put it before the Court.

The overwhelming volume of information available, in addition to the fact that much of it is duplicated, edited and even manipulated by partisan sources means that the credibility of OSINT material hinges on the methodology underpinning its discovery and verification. The pioneering work of the University of California, Berkeley is already paving the way for standardization of OSINT as evidence through the development of a protocol. As part of its efforts towards accountability for alleged international law violations in Yemen, GLAN has worked with open-source investigators Bellingcat and others to trial and implement a process aimed at increasing the weight afforded to OSINT in courts of law.  

These light-touch procedures guide the methodical extraction and preservation of key pieces of content from an otherwise unmanageable mass of information, with a view to increasing its evidential utility once it has been verified.

To ensure that the material is suitably preserved for use in court, RightsLab, HURIDOCS and Mnemonic have partnered with GLAN to develop a combined evidence preservation and management system. This integrates Uwazi, (open-source software designed by HURIDOCS for building and sharing document collections) with Digital Evidence Vault, a software tool that simplifies the extraction and preservation of digital content from websites and uses distributed ledger technology to timestamp everything that is collected. This process ensures that the first step in a chain of custody is established once the material is brought into the system.  

A related and pervasive challenge for the OSINT community is the time-sensitive threat of take-downs by the algorithms operated by YouTube and other platforms. Our system allows researchers to extract and preserve individual items of interest found online, while others like Yemeni Archive have the ability to preserve on a mass scale.

Although OSINT has long been a feature of powerful Amnesty International investigations and has also been used in the courtroom, litigators anticipate judicial skepticism. However, this need not be inevitable if the discovery, preservation and verification processes are considered by analogy to established concepts. Judges are already intimately familiar with the chain-of-custody principles that are being addressed by technologists. The need to authenticate documents is an integral feature of traditional litigation; the requirement to exclude evidence that is so unreliable as to be unfair to a defendant is the very reason we deprecate hearsay evidence. When OSINT analysts pore over a video, geolocating, cross referencing and reverse-image searching its contents, they are effectively checking its reliability. If an investigator discovers incriminating evidence by following an impartial, replicable procedure, preserves that evidence properly and analyses it to assess its authenticity, a convincing case can be made for its inclusion as evidence.

The use of OSINT does not have to mean driving accountability efforts out of the hands of Yemenis and into those of international investigators. Its analysis is far better carried out by, or involving, those with local knowledge, which is where collaboration platforms like Uwazi can help. It will take time before the full benefits of these steps (generously funded in this case by the Roddick Foundation, Nesta and Cherish-DE), are seen – but laying the foundations now is the small part we are playing in the pushback against impunity.