From venture-backed startups to AMD, Nvidia, and IBM, the tech-arms fusion is automating warfare — and squeezing human oversight out of the kill chain.
Source material: privacyinternational.org
The companies profiled by PAX and Privacy International map the fusion of the technology and arms sectors into a new military-industrial complex. The report covers chipmakers AMD, Cisco, IBM and Nvidia; cloud and software giants Alphabet, Amazon, Meta, Microsoft, Oracle and SpaceX; venture-backed Anduril and Palantir; and the five largest arms producers BAE Systems, General Dynamics, Lockheed Martin, Northrop Grumman and RTX. The selection is deliberately weighted to the biggest players and not exhaustive, but it documents a shared set of concerns: AI decision-support systems that compress the kill chain, the erosion of companies' own ethical commitments, and a dangerous concentration of power in a handful of firms with unusually close ties to government. The report argues these developments are outpacing the rules meant to govern them, and calls on states and companies alike to adopt binding international rules grounded in international humanitarian law and international human rights law, covering AI in the military domain and corporate accountability.
Two companies stand out in the new defense economy: Anduril and Palantir, both backed by venture capital. The report groups them as 'neo-primes' — a new breed of contractor that bypasses the traditional arms-industry path. Anduril builds autonomous systems, while Palantir supplies data analytics and AI tools to military clients. Their rise reflects a broader shift in which tech startups, rather than established defense firms, are shaping how AI gets used in conflict. The report traces how these companies have become key suppliers to the military, alongside the cloud giants and chipmakers. What distinguishes them is their origins in the commercial tech sector and their willingness to move fast, unencumbered by the Pentagon's legacy procurement processes. Both have drawn scrutiny for the ethical implications of their work, particularly the automation of target selection and surveillance. Their prominence signals that the boundary between consumer tech and warfare is eroding.
The hardware underneath modern warfare comes from four companies named in the report: AMD, Cisco, IBM and Nvidia. Their chips and networks form the computing backbone for military AI systems, from data centres to battlefield sensors. Nvidia's GPUs, for example, are widely used for training machine-learning models, while Cisco's networking gear carries the data that feeds them. The report identifies these hardware producers as critical to the militarisation of AI, noting that their components are essential to the precision and autonomy of next-generation weapons. While these companies may not design weapons themselves, their products enable the software that does. The report examines their military contracts as part of a larger pattern of convergence between the tech supply chain and the arms industry. The dependence of modern militaries on commercial off-the-shelf chips creates a new kind of vulnerability: a single company's decision to restrict exports can have immediate consequences for military capability.
AI decision-support systems are squashing the kill chain, the report finds. The kill chain is the sequence of steps from detecting a target to engaging it, and AI is compressing it to near-instantaneous speed. This compression weakens meaningful human control, because operators no longer have time to review the output of algorithms. As systems become more autonomous, the human role is reduced to rubber-stamping decisions made by machine-learning models. Errors become harder to catch. The report calls for binding international rules that require human oversight at every step of targeting. It notes that the technology is already being battle-tested. Without new standards, the principle of distinction under international humanitarian law will be impossible to uphold. The only solution is to design systems that keep humans in the loop.
Anduril and Palantir, the venture-backed 'neo-primes' identified by PAX and Privacy International, represent a new kind of military contractor. They are not part of the traditional arms industry, nor are they the largest tech firms; they are a distinct category that fuses software development with weapons production. Their business model relies on close ties to government, contributing to the dangerous concentration of power that the report traces across all profiled companies. They also share in the erosion of ethical red lines, as they deploy AI decision-support systems that compress the kill chain and undermine meaningful human control. With military budgets expanding, these neo-primes are set to shape the future of warfare. Any binding rules for AI in the military domain must apply to them, and to the corporate accountability that their activities demand.
The PAX and Privacy International report documents that the same technologies that power military AI also drive mass surveillance, eroding privacy both in war and in the domestic surveillance state. Cloud platforms, data analytics and machine learning are used to pick out targets for strikes and to monitor civilian populations, blurring the line between military and civilian applications. The report argues that this erosion of privacy is not a side effect but a central feature of the new military-industrial complex, as companies sell the same tools to defence and law enforcement agencies. The concentration of power in a few tech giants, with their close government ties, means that decisions about who is watched and who is targeted rest in the hands of a small number of unaccountable firms. The report's remedy is to apply international human rights law strictly, ensuring that surveillance is lawful, necessary and proportionate, and to hold corporations liable for abuses. Yet the lack of transparency around military contracts prevents citizens from discovering how their data is actually used, making democratic oversight nearly impossible.
The PAX and Privacy International report identifies civilian digital infrastructure as a dual-use domain: it is both a weapon and a target. The same undersea cables, cloud services and communication networks that sustain civilian life can be used for military operations, and are vulnerable to attack from adversaries seeking to paralyse a state. This intertwining means that commercial infrastructure deployed in battle zones effectively becomes a military asset, as the report notes. Yet the legal framework designed to protect civilian objects lags far behind. The report argues that without explicit protection for digital infrastructure, the next conflict could see the internet itself become a casualty. It therefore calls for binding international rules, grounded in humanitarian law, to cover these systems, and for states to hold companies accountable when their infrastructure is used to cause disproportionate harm. The challenge is that the same networks that enable mass surveillance and military AI also provide essential services to civilians, making distinction nearly impossible.
Where this came from. This breakdown is based on source material published at privacyinternational.org. Images above are used with the credits shown beneath each one.
