At Rise of AI 2024 in Berlin, a panel with Yvonne Hofstetter (CEO, 21strategies), Lieutenant General Michael Vetter (CIO, German Federal Ministry of Defense) and Ulrike Franke (Senior Policy Fellow, European Council on Foreign Relations), moderated by Katja Munoz, discusses how Germany can not just keep up but thrive in AI-driven warfare. Using a moonshot-style format, the panel first reviews the status quo: Vetter describes the Bundeswehr's new digitization platform and projects such as the digitized battlefield, an electronic-intelligence AI laboratory and a common operational picture; Hofstetter argues defense AI has changed little in 25 years and that information dominance alone is not enough, citing Ukraine; Franke says the Bundeswehr has been underfunded for decades and is now rebuilding everything at once, making a formal military AI strategy less urgent than actually building capability. The panel then sets two- and five-year visions covering drone swarms, autonomy rules, procedures for AI targeting recommendations, software-defined defense, and Germany's Defense Metaverse test environment, before naming what each considers most necessary to achieve them.
Key takeaways
- Vetter says the Bundeswehr reorganized IT procurement into nine NATO C3-based clusters, including one for AI analytics and simulation.
- Hofstetter argues defense AI has changed little in 25 years and that information dominance alone does not decide outcomes, as Ukraine shows.
- Franke identifies three layers of uncertainty in military AI: the technology itself, its applications, and how it ends up being used.
- Franke cites unconfirmed reports of Israeli AI targeting systems "the gospel" and "lavender" in Gaza; moderator Katja Munoz says a reported two-second human review would not be meaningful human control.
- Hofstetter describes Germany's Defense Metaverse, Ghost Play, as a unique testbed that lets allies and Germany evaluate AI systems before deployment.
Detailed notes
Status quo: the Bundeswehr's digitization platform
Vetter says Germany faces a Zeitenwende in security policy that coincides with a Zeitenwende in technology, citing US General Millie's description of a seismic change in the character of war and pointing to the ubiquity of sensors and loitering munitions in Ukraine. He describes a digitization platform launched three years ago that reorganized Bundeswehr IT procurement into nine clusters based on NATO's C3 taxonomy, including one for analytics and simulation. Current AI projects include a digitized battlefield using pattern and object recognition and sensor fusion to pursue information superiority, an AI laboratory for electronic intelligence performing anomaly detection, and a common operational picture fusing standardized and unstandardized data for the cyber and information domain.
Twenty-five years of defense AI and "steel at the battlefield"
Hofstetter recalls that her CTO helped build an early defense AI system in 1997 for the AWACS airborne early-warning system, fusing multiple data sources into a recognized air picture, and says the field has since made less progress than expected: newer startups often rediscover problems already solved twenty years earlier. She recounts a general at that 1997 briefing insisting that "all that counts is steel at the battlefield," a claim she says looking at Ukraine now, she believes was right; information dominance alone is not enough once systems must translate into embedded AI on physical weapon platforms and logistics. She predicts AI will move beyond the information and cyber space into what she calls ambient intelligence integrated with physical platforms.
Underfunding and the missing German military AI strategy
Franke says the Bundeswehr was significantly underfunded for decades, leaving no room to experiment with new capabilities, and that the current 100-billion-euro special fund is being used to rebuild many capabilities simultaneously, including manned aircraft like the F-35 alongside newer AI and autonomy investments, forcing hard choices. She argues Germany's long-criticized lack of a dedicated military AI strategy, unlike France, the US and China, mattered five years ago but is now less important than simply building capability; Vetter adds that of the 100-billion fund, 21 billion euros goes to digitization and command and control, because without it forces cannot lead formations even with strong hardware like the Leopard 2 tank. Vetter says he prefers delivering a 40 percent solution to soldiers quickly over a fully gold-plated strategy delivered late; Hofstetter agrees a strategy is useful mainly as a stringent, regularly rebalanced portfolio of options with assigned costs.
Two- and five-year visions
Franke identifies three layers of uncertainty limiting predictions: the technology itself is still fast-moving, its military applications are unclear, and how it is used matters more than the technology, as with tanks in the World Wars. She prioritizes drone and other swarming capability, defining rules for where to use autonomous systems, and developing procedures for AI-generated targeting recommendations, citing unconfirmed reports of Israeli AI systems named "the gospel" and "lavender" in Gaza; moderator Katja Munoz adds that a roughly two-second human review reported for one of these systems would not amount to meaningful human control. Vetter counters that some defensive systems, such as the German Patriot missile system and the naval RAM system, already operate in highly automated mode because there is no time for human deliberation, distinguishing that from offensive lethal autonomous weapons, which he calls a different, highly ethical question. His five-year vision is a "software-defined defense" in which an application layer is added to legacy platforms such as the Eurofighter, alongside a constant DevSecOps-style cycle of test, evaluation and rollout, plus cultural and training change. Hofstetter's vision centers on deterrence rather than costly, vulnerable prototype technology, achieved over five years.
Germany's Defense Metaverse
Hofstetter highlights the Defense Metaverse, working title Ghost Play, initiated by the Ministry of Defense around 2020-2021, which lets Germany test AI systems and revise their specification before real-world deployment, including checking for ethical behavior. She says NATO allies including the UK, Sweden, Denmark and Spain have asked about it after demonstrations in Washington, including one attended by the US Department of Defense and intelligence services. She cites a Helmut Schmidt University study of 30 countries' defense AI programs finding most cluster around intelligence, surveillance and reconnaissance, with Germany and Turkey as outliers, Turkey betting heavily on full autonomy in weapon systems.
Q&A highlights: what each panelist would prioritize
Asked for the single most necessary step, Hofstetter says industry must show more ambition after decades of relying on modest Ministry of Defense maintenance funding. Franke says no single answer suffices; Europe still lacks real urgency about how fast the geopolitical and technological situation is changing, and needs money plus genuine willingness to work with the private sector as a whole-of-society effort. Vetter's closing vision reiterates that culture and people, not only technology, must change alongside the digital backbone for the Bundeswehr to participate in multi-domain operations within five to six years.
Quotes
Looking at Ukraine, I think we did wrong to that general, because he was right: all that counts is steel at the battlefield.
Yvonne Hofstetter
If you just have a human that then presses the button because the AI told them to, then you might as well not have the human.
Katja Munoz
I'm pretty much more in favor of delivering capability to the war fighter, not a 100% gold-plated solution but a 40% solution that he gets tomorrow.
Michael Vetter
Questions answered in this talk
Why doesn't Germany have a dedicated military AI strategy?
According to Ulrike Franke at Rise of AI 2024, having a written strategy would have been useful five years ago, but the Bundeswehr was underfunded for decades and is now rebuilding many capabilities at once, so the priority has shifted from writing strategy to actually building capability.
What is Germany's Defense Metaverse?
At Rise of AI 2024, Yvonne Hofstetter describes it as a German military testbed, working title Ghost Play, initiated around 2020-2021, that lets AI systems be tested and their behavior, including ethical conduct, checked and revised before real-world deployment; NATO allies such as the UK, Sweden, Denmark and Spain have asked about it.
Should AI-generated military targeting recommendations be trusted without close human review?
At Rise of AI 2024, panelist Ulrike Franke cited unconfirmed reports of Israeli AI targeting systems, "the gospel" and "lavender", used in Gaza and called for clear rules on using such recommendations; moderator Katja Munoz added that a roughly two-second human review reported for one of these systems would not amount to meaningful human control.