02 Jul Upstream Harms: OSI, AI Militarism, Big Tech, and Gaza
[Patrick Brian Smith is an Assistant Professor and University Fellow at the University of Salford, where his research critically examines how evidence is mediated with a particular focus on open-source investigation, AI, and human rights]
Human Rights-focused OSI
OSI (open source investigation) practices involve the collection and analysis of information drawn from publicly accessible sources to examine specific events, individuals, and organisations. Although OSI practices predate the digital era, within:
“our digitally networked, mediated, and platformed present, the sheer ubiquity of—and access to—diverse forms of media evidence”
has profoundly reshaped both the uses and methodologies of OSI. OSI is playing an increasingly important role in human rights investigations and activism. Investigators and advocates are increasingly using online evidence to document abuses and support accountability efforts. Where investigations once relied primarily on on-the-ground documentation, much of this evidence now emerges from digital spaces.
OSI practices have been key to human rights-focused investigations into the ongoing killing and destruction in Gaza (which the International Court of Justice has ruled as a plausible genocide), particularly as Israel has maintained a near-total ban on independent foreign journalists and international human rights investigators entering the territory. To date, most of this OSI work has focused on retrospectively documenting civilian deaths, destruction, and human rights violations. For example, a statistical study published in The Lancet estimated:
“75200 violent deaths (95% CI 63 600–86 800) between Oct 7, 2023, and Jan 5, 2025, representing approximately 3.4% of the Gaza Strip’s pre-conflict population.”
The scale of killing in Gaza has been rapidly accelerated through the utilisation of AI targeting systems employed by the Israeli military. Investigative reports from late 2023 to early 2024 detailed several AI-driven systems (‘Habsora,’ ‘Lavender,’ and ‘Where’s Daddy?’) that were being used by the Israel Defence Forces (IDF) for target identification. Such systems “can ‘generate’ targets almost automatically… essentially facilitat[ing] a ‘mass assassination factory.’” The automation and acceleration of target-generation processes have compressed kill chains, with AI now embedded across military systems. Operating across surveillance, targeting, logistics, and battlefield management, AI is increasingly directly influencing the conduct of hostilities.
The imbrication of big tech companies in this emerging ecosystem is increasingly obvious. In 2021, Israel announced Project Nimbus (PN), a $1.2bil cloud partnership with US tech giants Google and Amazon Web Services (AWS) to provide cloud computing for the government and military. This included the construction of multiple data centre facilities in Israel to support the state’s private cloud migration, effectively representing a partnership between the heavily militarised Israeli state and two of the world’s three largest cloud infrastructure providers. Here then, the socio-technical systems of big tech companies and their technologies bracketed under the term ‘artificial intelligence’ (AI) are increasingly intertwined with the modalities of modern imperialist warfare.
The downstream violences enabled by these AI systems are clear to see and have been well-documented. Here, however, I want to argue that OSI practices could, and should, be refocused to better interrogate the upstream, material infrastructures that undergird these techno-militaristic systems and, by extension, the complicity of big tech organisations therein. By reorienting OSINT from a largely reactive practice of documenting destruction and killing after it occurs toward a proactive methodology focused on the infrastructures, institutions, and supply chains that enable Israeli state violence, investigators can intervene earlier and more effectively within these techno-military ecosystems.
As Tor Krever has argued, international law has been unable to effectively “grapple with the material structures and systemic logics out of which violence and atrocity arise.” A retooled and refocused OSI practice—centered around an examination of the corporate, logistical, and infrastructural networks that sustain military AI systems through the analysis of evidence such as government contracts, procurement records, corporate disclosures, investor reports, cloud infrastructure documentation, and environmental permitting records, amongst other elements—could effectively respond to the core failure identified by Krever. Rather than focusing on identifying and reconstructing acute episodes of military violence, such investigatory foci would help to reveal the broader assemblages of actors, infrastructures, and resources through which military AI capabilities are produced and maintained.
By identifying these relationships upstream, before downstream violence occurs, OSI can carve open a space for more effective public advocacy, labour organising, regulatory intervention, or legal scrutiny aimed at disrupting these harmful infrastructures before they become entrenched. As Krever continues to suggest, international law:
“has a tendency to focus attention on particular incidents and outbreaks of violence without ever systematically engaging with underlying structural forces.”
Such a retooled approach to OSI can, I argue, help push legal accountability frameworks outside their comfort zone. Expanding accountability beyond discrete incidents and visible acts of violence means we can more effectively resist the structural, systemic, and often obscured forms of harm embedded within the development, deployment, and maintenance of military AI systems. In doing so, human rights OSI can move beyond evidentiary reconstruction toward anticipatory intervention—identifying emerging risks, exposing relationships of complicity before abuses occur, and creating opportunities for preventative action—whilst also broadening the scope of what accountability itself might entail.
Project Nimbus and the Infrastructural Complicity of Big Tech
PN provides “critical infrastructure that supports data storage, machine learning, and AI applications.” At the time of the project’s announcement, the ministry described PN as an “all-encompassing cloud solution” designed to serve “the government, the defense establishment, and others.” However, Google stated in early 2024 that PN is intended only for civilian workloads, explicitly excluding military applications related to weapons or intelligence. Despite these assurances, investigative reports have increasingly highlighted connections between PN and the IDF. Caroline Haskins found that the IDF have been centrally involved with PN from the outset, shaping its design and utilising it operationally—a link underscored by Israel’s cyber chief Gaby Portnoy, who publicly attributed battlefield successes to PN. Abraham further revealed how the IDF, whose storage and processing servers became overwhelmed post-October 7th, 2023, turned to PN’s cloud infrastructure for additional power, emphasising the operational benefits of its extensive AI capabilities, which enhanced their effectiveness in Gaza.
It is important to remember that these interrelations are underpinned by the broader “transition of major… tech companies into defense contracts.” In the US context, we have seen the deepening interconnections between Silicon Valley and the U.S. military–intelligence establishment. Research has examined how Pentagon initiatives like the Defense Innovation Unit and the CIA’s venture capital arm have channelled billions into major tech giants. This:
As Chua et al. write about Israel’s contracts with big tech companies, campaigners have been warning of the militarisation of the US tech industry for at least the past decade. Within this landscape, big tech increasingly functions as an extension of imperialist interests.
Griffiths and Rubaii have put forward the notion of ecological “beforemaths” to highlight how the relationship between war and the geos—the “the ground and the life it sustains”—is doubly destructive. For them, significant attention has been paid to the ecological aftermaths of war—those “toxic materials [that] threaten life… as… deposits of bombardment”. However, they emphasise the need to refocus on the pre-war sites of extraction “that make advanced military technologies possible.” They examine how this shift in focus might broaden understandings of war to encompass its spatial and temporal dimensions, as well as the diverse range of bodies it affects. This reorientation forces us to more capaciously account for the range of human and environmental bodies implicated and harmed across the global systems that enable warfare. I argue that their methodology can be effectively taken up to analyse the upstream, infrastructural underpinnings of the developing big tech-military ecosystem in Israel. By exposing the distributed infrastructures through which military AI systems are produced and maintained, OSI can help to identify intervention points where advocacy, legal action, labour resistance, divestment campaigns, or regulatory measures might be applied. Again, this can allow for a more proactive and preventative OSI methodology to be developed, focused more concretely on the anticipatory exposure of the increasing roles of big tech companies in potential human rights abuses.
Kotliar and Gekker have examined how PN centrally “revolves around the construction… of six massive, resource-hungry corporate data centers.” The continued occupation of—and concomitant resource extraction from—the Palestinian territories are likely to be central to the maintenance of these new data centres. It is impossible to separate the western protection of Israeli interests in Palestine and the region more broadly from its role in securing these lands and ecologies for capital. As Bethan McKernan explains:
“Israel controls about 80% of water reserves in the West Bank, but both the West Bank and Gaza Strip face severe water stress and drought.”
The data centres that support Israel’s tech-defense contracts require large amounts of water, with larger infrastructures using millions of gallons per day. In Israel, AWS has three “availability zones” spread across three AWS data centers, located in Tnuvot (Tnuvot industrial zone), Shoham (Shoham industrial zone, near Ben Gurion Airport), and Beit Shemesh (Har Tuv A industrial zone). Collectively, these are known as ‘il-central-1 AWS region: Middle East (Israel).’ The centers became operational largely because of 4AWS (alongside Google) winning the PN tender. All three centers are approximately 5km from the border with the occupied West Bank.
There is evidence that both the Shoham industrial zone and Har Tuv A industrial zone are located near wellfields (Rosh Ha’ayin/Yarkon and Kfar Uriya and Eshtaol) that pump from the Yarkon-Taninim (Western Mountain) Aquifer. The Western Aquifer’s recharge is predominantly on the West Bank highlands (about ~73% of recharge occurs there), flowing west to Israeli pumping centers. Thus, there is strong evidence that two of the three areas where these data centers are located tap a transboundary aquifer that is recharged largely in the West Bank. Access to water becomes a “potent state-controlled weapon” to maintain and extend the Occupation. The expansion of Israel is not only a military imperative driven by ideological zealots, but also one that the future economic growth of the country depends on, as with other settler colonial states.
Due to their effective monopoly over cloud computing, with AWS, Google, and Microsoft controlling over 60% of global market share in cloud services in 2024, US big tech is increasingly at the centre of twin transition policies. In this sense, Israel is operating as an engine of wider technologies of imperialist expansion in Western Asia and the Mediterranean that require the effective integration of territories as either engines of—or extractive zones for—energy security in an unstable geopolitical milieu. Widely reported gas reserves off the coast of Gaza, as well as aforementioned water and land resources utilised towards data centres to support the Israeli state and its securing of interests in the region, are all embedded in a political ecology of empire within which the electricity requirements of large-scale computing are increasingly central.
With the deepening of the role of big tech and its infrastructural resource management paradigms into the landscape of militarisation, from the US to Gaza, what we are seeing is a gear shift in the already-existing systemic function of US tech companies in the maintenance of the interests of the US empire. OSI strategies must be better attuned to uncovering the infrastructural interconnections between big tech and military actors. Palestine has been widely treated as a testing laboratory for Israeli military techniques and technologies, and is frequently regarded as an “exception” in the application (and enforcement) of international human rights frameworks. This dual logic—of experimentation and impunity—accelerates both violent forms of technological innovation and broader forms of domination.
Conclusion
The task for human rights OSI, then, is not only to document or reconstruct the chain of events once the harm has already been done. It must more effectively and holistically move upstream, mapping the infrastructures, institutions, resource regimes, corporate partnerships, and logistical systems that make such violence possible. In the context of Israel’s assault on Gaza, this means investigating not only military actors but also the cloud architectures, data centres, extraction networks, and corporate contracts that sustain an increasingly automated machinery of occupation and mass violence. The devastation of Gaza is not the product of isolated decisions or technological accidents; it is enabled by a vast socio-technical ecosystem whose participants frequently remain shielded from scrutiny.
A proactive OSI practice must therefore seek to expose these relationships before they culminate in further atrocities, transforming investigation from an act of retrospective witnessing into a tool of prevention, intervention, and resistance. Of course, exposure alone is unlikely to play a significant role in the prevention of atrocities. The value of this retooled form of OSI praxis lies in its ability to expand the very temporal and political horizons of accountability. Exposing enabling infrastructures before violence occurs can help to foster new opportunities for intervention, disruption, and accountability upstream of harm. If Gaza reveals anything, it is that accountability cannot stop at the trigger. It must extend to every actor, institution, and infrastructure that helps make such violence and destruction possible. Gaza, and the wider landscape of the Israeli occupation, is a snapshot of a future we are hurtling towards. At the same time, the principled resistance to these technologies is rising, in Palestine and beyond. It is an existential imperative to do all we can to support them.
Leave a Reply