At the Rise of AI Summit 2020, Dr. Ulrike Franke, a policy fellow at the European Council on Foreign Relations, argues that military AI deserves more public attention in Europe than it currently gets. She makes four points: AI is already used across intelligence analysis, logistics, cyber operations, command and control and drone swarming, with roughly two-thirds of US Department of Defense AI-related investment going into AI-enabled systems; civilian and military AI are not cleanly separable, as shown by the 2017 controversy over Project Maven, a US Pentagon-Google cooperation on drone footage analysis; AI could fundamentally affect warfare and global stability, including risks to nuclear deterrence if AI-enabled sensors make oceans 'transparent' to submarine detection, and risks of rapid 'flash wars' between interacting autonomous systems; and Europe's political sphere remains reluctant to discuss military AI publicly, with France as the notable exception. She closes a Q&A on AI proliferation to non-state actors and the drone-heavy conflict in Nagorno-Karabakh by calling for European governments to discuss military AI more openly.
Key takeaways
- Franke notes roughly two-thirds of US Department of Defense AI-related investment goes into AI-enabled systems, plus further funding for autonomy.
- She warns AI-enabled sensors could make oceans 'transparent,' threatening the second-strike capability underpinning nuclear deterrence.
- Franke describes the risk of 'flash wars,' where interacting autonomous military systems could escalate an incident within seconds, similar to financial flash crashes.
- Of 21 national AI strategies published by EU countries, only around five mention military AI, and Germany's contains just one sentence on defense.
- France is the only European country with a dedicated military AI strategy, committing EUR100 million a year to defense AI research.
Detailed notes
Where AI already shapes military operations
Franke lists current military uses of AI: intelligence, surveillance and reconnaissance systems that sift through vast volumes of drone footage; logistics, including predictive maintenance such as on the next French Rafale fighter jet; cyber operations; command and control that helps combine intelligence for joint operations; swarming, such as the UK's 'many drones make light work' project coordinating up to 20 drones under one operator; and forecasting, such as a German Bundeswehr project using open data to anticipate crises. She notes roughly two-thirds of US Department of Defense AI-related investment goes into AI-enabled systems, with further funding for autonomy more broadly.
The blurred line between civilian and military AI
Franke argues civilian and military AI are not cleanly separable: drone swarms used as firework substitutes in China or to spell out Joe Biden's name after his 2020 election victory rely on the same technology as military swarming, and facial or language recognition built for civilian products can be repurposed for targeting. She cites reports that the CIA was investing in over 140 AI projects involving data analysis and facial recognition, and recounts the 2017 Project Maven controversy, in which Google employees walked out after learning the company was helping the Pentagon analyze drone footage with machine learning, ending Google's involvement even as broader Silicon Valley-military cooperation continued.
Risks to global stability: submarine deterrence and 'flash wars'
She warns that AI-enabled sensors could eventually make oceans 'transparent,' making it possible to locate submarines carrying nuclear weapons and undermining the second-strike capability that underpins current nuclear deterrence. She also describes the risk of 'flash wars,' a term borrowed from financial-market flash crashes, where interacting autonomous military systems could escalate a minor incident or misunderstanding into conflict within a very short time frame, with no side intending it.
Europe's reluctance to discuss military AI, and France's exception
Franke finds that of 21 national AI strategies published by EU member states, only around five mention military applications at all; Italy, NATO's fifth-largest defense vendor, does not address the topic, and Germany's strategy contains one sentence on defense. France is the exception, publishing a dedicated military AI strategy, naming security and defense as one of four priority AI industrial sectors, and committing EUR100 million a year to defense AI research; the Netherlands is following a similar path. She contrasts this with active European military AI programs already underway, including the Franco-German-Spanish Future Combat Air System (FCAS) and unmanned combat drone projects such as France's Neuron and the UK's Taranis.
Q&A highlights
On the risk of AI-driven terrorism, Franke notes that once an AI model is trained it is essentially software that can be shared, making diffusion to non-state actors a real concern for cyberattacks and disinformation, and says Europe is not well prepared for it. On the Nagorno-Karabakh conflict between Azerbaijan and Armenia, she characterizes it as primarily a drone war using largely off-the-shelf Israeli-bought systems rather than heavily AI-enabled ones, and predicts wealthier, better-equipped states will deploy AI-enabled military systems first before such capability eventually diffuses more broadly. Asked for a closing wish, she says she wants European governments to talk about military AI more openly rather than avoid the topic.
Quotes
Civilian artificial intelligence can be militarized and weaponized.
Ulrike Franke
As with every kind of military system, the moment it's being developed, someone tries to develop a counter system.
Ulrike Franke
I would really like the Europeans to just talk about this more openly, think this more clearly.
Ulrike Franke
Questions answered in this talk
How is AI already used in the military?
According to Dr. Ulrike Franke at the Rise of AI Summit 2020, militaries already use AI for analyzing drone footage, predictive maintenance and logistics, cyber operations, command and control, and drone swarming, with autonomous systems taking a growing share of investment, such as roughly two-thirds of US Department of Defense AI-related spending.
Could AI undermine nuclear deterrence?
Franke warns AI-enabled sensors could eventually make oceans 'transparent,' making it possible to locate nuclear submarines and undermine the second-strike capability that is a core pillar of current nuclear deterrence.
Why doesn't Europe talk publicly about military AI?
Franke finds only around five of the 21 national AI strategies published by EU member states mention military applications at all, with France the main exception, having published its own military AI strategy and committed EUR100 million a year to defense AI research.