Details
Triggered by the many wars that are being waged at the moment, an arms boom has created a new ecosystem of defence start-ups, data science, artificial intelligence, and drone technology. With the rollout of AI-based (semi-)autonomous warfare, a vast new technology market has sprung up almost overnight. And drone start-ups have not only emerged in Ukraine. In the last few years, successful companies involved in the development and production of drones have also appeared in Turkey, Iran, and in Russia.
The EU and Germany, grappling with multiple crises, now also want to benefit from this new defence tech market. Accordingly, an EU white paper calls for “an ecosystem of technological innovation for our defence industries.” It is often unclear exactly what AI means in the military systems sector or what the work of these start-ups is based on. At the moment, AI mostly comprises the processing of large quantities of heterogeneous data (big data) with the help of machine learning and high-performance computers.
Processing data involves formalized data analysis with the help of statistics and subjective probabilities, embedded in a vast infrastructure made up of energy-gobbling data centres and the human labour required to train the algorithms. The data mostly comes from large surveillance platforms owned by companies such as Meta, Google, Amazon, and Microsoft. Pattern and image recognition systems and large language models (LLMs) have high error rates, for example because LLMs have no real understanding of the content they process, and reproduce the prejudices of the data they have hoovered up.
Despite the high error rates, these technologies are used in war contexts for reconnaissance, drawing up kill lists, highly-automated targeting and decision-making, and autonomous weapons systems.
Defence Tech Unicorns
Defence giants like Rheinmetall and Diehl, which primarily specialize in hardware, arms, and transatlantic operations, have not been the principal players in developing these “ecosystems” of data science, artificial intelligence, drones, and other software and AI-based technologies. The most sought-after products are instead made by US AI and software-centric start-ups like Anduril, Shield AI, and Palantir, the French AI start-up Mistral, and German firms like Helsing, Stark, Arx Robotics, and Quantum Systems, which, save for Palantir, are all newcomers to the arms industry. Software companies like Helsing entered the AI-powered drone business only a few years ago and have since developed into one of the “defence tech unicorns” — start-ups with a valuation of over a billion US dollars. The valuation of the military AI market was estimated at over 13 billion USD in 2024; in the next seven years this figure should swell to 35 billion USD.
Previously, start-ups had primarily operated in the civilian sector, as the military sector had been seen as a consolidated field with plenty of red tape, negligible profit margins, and dominated by a small number of industry giants. However, with the rapidly accelerating advent of algorithmic warfare, AI-based technologies are now also being presented as a game changer that will upend the old military-industrial complex. The new start-ups loudly proclaim their “disruptive” technologies and products as well as their limitless capabilities for extracting venture capital — even with products that are anything but market-ready. In the summer of 2025, Ukraine’s Ministry of Defence complained that Helsing drones were overpriced and did not live up to their promised capabilities.
Move Fast and Break Things
At the same time, military start-ups are happy to challenge established regulations and long-term procurement processes under the pretext of dismantling bureaucracy, while in reality they only seek to circumvent regulatory oversight. Palmer Luckey, CEO of Anduril (the largest defence tech start-up), boasted that his company spends more on lobbyists and lawyers than engineers. There is a burgeoning culture of rule-breaking with impunity — not only in the US, but also in Germany. Hence, in the summer of 2025, neither the announcement of supplying drones nor the building of munitions factories in Ukraine faced any scrutiny from the mainstream media or parliament when it came to the relevant approval processes.
These developments align with the well-known Silicon Valley credo, “move fast and break things”. The aim is to speed up economic, bureaucratic, and technical processes. Bloated by venture capital, firms buy up their competitors in order to expand capabilities, eliminate competition, and create large hubs or monopolies as quickly as possible. These firms are attempting to position themselves as key actors with comprehensive capabilities, so as to carve out the biggest possible piece of the pie of the newly-announced defence budgets.
The short cycle of innovation in the disruptive AI defence tech sector results in an experimental form of warfare, where any mistakes are paid for with the blood of civilians and soldiers alike.
Whether these capabilities are real is hardly ever investigated. For example, in September 2025 the software and drone firm Helsing presented the prototype of a stealth bomber that Helsing’s own PR claimed had been designed in as little as 14 days. Just a few weeks prior, Helsing had acquired the small Allgäu-based aircraft manufacturer Grob Aircraft, which is now serving as a platform for the development of the first German drone bomber.
This approach is similar to the core business model of the young defence tech start-ups: in software development, extremely short production cycles are the rule; the Helsing drones deployed in Ukraine are provided with software updates every week. No one can say with certainty what effects speeding up will end up having. It is unclear whether, and by whom, such updates are reviewed. While it is a common, albeit questionable, practice in the civilian sector to have users serve as unpaid product-testers for half-finished software, the short cycle of innovation in the disruptive AI defence tech sector results in an experimental form of warfare, where any mistakes are paid for with the blood of civilians and soldiers alike.
Defending Democracy or Ending The World
In 1961, US president Eisenhower — himself a former general — warned of the influence of the military-industrial complex and that it could undermine democracy. But mutual influence and interdependence are increasing in the age of AI-based warfare. Many start-up CEOs enjoy close links to politics, military, and business. Helsing CEO Gundbert Scherf was a consultant for former German defence minister Ursula von der Leyen. US start-ups are often closely tied to authoritarian venture capitalists in Trump’s orbit such as Peter Thiel and Elon Musk. At the same time, military start-ups consciously boost already completely overinflated expectations of AI-assisted military systems.
By doing so, they aim to secure the availability of further state resources for AI. In a New York Times article, Alex Karp, CEO of the security and AI-weapons platform Palantir, spoke of AI as an “Oppenheimer moment”. Robert Oppenheimer is considered the father of the atomic bomb. According to Karp, the development of AI will be equally impactful. Here, salvation and damnation often form part of the same story.
Among start-ups, “defending democracy” is a favoured narrative.
Eric Schmidt, former CEO of Google, as well as a financial backer of Rebellion Defense and owner of defence start-up White Stork, thinks that “we” have to develop AI faster than adversaries such as China or Russia, in order to protect Western democracies and the “free world” from authoritarianism. This is also reflected in the 2021 report of the US National Security Commission for Artificial Intelligence, which Schmidt chaired.
Among start-ups, “defending democracy” is a favoured narrative; as for example with the German drone start-up Stark, where this means nothing more than wanting to sell their own weapons to NATO countries. A rather feeble declaration, not only in the face of the erosion of the transatlantic partnership, but also considering the autocratic tendencies among NATO partners Turkey, Romania, and the US.
A Drone Wall to “Defend Democracy”
“Defending democracy” is also a concern for Helsing’s Gundbert Scherf who in early 2025 proposed creating a “drone wall” on NATO’s eastern flank to protect Western democracy. The wall would be comprised of tens of thousands of reconnaissance and kamikaze drones that could form a swarm with the help of AI. Scherf proposed this clever business idea shortly after Germany softened its debt brake for defence spending, and NATO defined a new defence spending goal of five percent of GDP.
Scherf sold the idea with the following rationale: “autonomous systems are made for democracies. We value life, we all like to live a good life. I don't believe that our democracies can or want to wage a war of attrition that costs many lives … That means we really need to have these asymmetric technological capabilities. Asymmetric technology wins!” Scherf is suggesting that democracies need autonomous decision-making, targeting, and weapons systems, as in post-heroic times the local population will not submit to conscription. Technological superiority alone, he claims, will enable us to combat the multiple crises of our age and win future wars. Why the US ended up losing the wars in Vietnam and Afghanistan, if asymmetric technology — that even back then was based on data — always “wins” remains a mystery.
The idea of the drone wall is pure technological solutionism, an ideology that has long been propounded in Silicon Valley and that encourages us to believe that technology alone can solve all our problems.
The idea of the drone wall is pure technological solutionism, an ideology that has long been propounded in Silicon Valley and that encourages us to believe that technology alone can solve all our problems. For European politics, AI-based tech and organized rule-breaking also seem to hold far more promise than negotiations, cooperation, development work, or social-ecological solutions.
Military build-up is increasingly regarded as the last lifeline for the eroding Western world order, overlooking the fact that such build-up completely goes against the Paris climate goals and will stoke further war. The continuing rearmament fever in Germany is especially irritating if one recalls that German civil society has debated the human rights and ethical issues around the development and use of drones for over ten years. Despite the fact that the Greens have long criticized the creation and arming of drones, it was the red-green-yellow coalition government [of the SPD, Greens, and FDP] that allowed the arming of remote-controlled, non-autonomous drones.
In July 2025 the German Bundeswehr acquired (semi-)autonomous drones “on a trial basis” — likely Helsing’s HX-2 drone and Stark’s OWE-V — describing them as munitions or ”loitering munitions”. This entails equipping drones with explosives to create fully or semi-autonomous so-called “kamikaze” drones that can “hang” in the air for hours, before striking a target in order to kill or destroy it. The decision to acquire these drones was likely made when Germany was in transition between governments in early 2025. There were many reports on the Bundeswehr’s plans to use suicide drones in the German media, but little information from the ministry of defence itself.
Who Pulls the Trigger?
The political process that led to the acquisition appears to have amounted to a serial breach of democratic norms. The Bundeswehr investing in drone technology was not debated in parliament, and it was hardly discussed in civil society either. There was a complete lack of critique in the mainstream media. Those responsible rushed to assure the public that these systems would always be connected to a human (“human-in-the-loop”) and that soldiers would be able to stop attacks at any moment.
The Geneva Conventions state that every military action must be the responsibility of a human actor. With autonomous systems, that is no longer possible.
Nevertheless, one could read on Stark’s website: “The AI-supported software for Virtus reacts to events in real time … The system can track and engage precise targets up to 100km away — even in low-signal or impaired operational environments.” Without network signal to the drone operator, only autonomous drones would be able to strike their targets.
The Geneva Conventions state that every military action must be the responsibility of a human actor. With autonomous systems, that is no longer possible — a “human-in-the-loop” would be of no help, as they cannot make a meaningful decision if they are not as fast as the machine. Thus life-and-death decisions are left to hastily designed algorithms and systems. The disruptive AI military-industrial complex, then, does not “defend freedom”, it undermines democracy and international law.
This article was first published in German in the journal LuXemburg
Translated by Gegensatz Collective
