AI in Defence in Europe: Innovation, caution and the push for human control

AI in defence is no longer just a policy slogan. Under the 2024 European Defence Fund (EDF), the European Commission awarded grants to 62 collaborative defence research and development projects and began signing the related grant agreements in December 2025. These projects involve more than 600 legal entities from EU Member States and Norway, with an EU contribution close to €1 billion. They cover areas such as air and missile defence, drones and counter-drone systems, artificial intelligence, quantum technologies and military mobility. In April 2026, the Commission announced a further €1.07 billion for 57 projects selected under the 2025 EDF.

Several projects show how concrete these developments have become. NEUROQUAD combines neurotechnology, AI and quantum computing to monitor and predict the cognitive state of military pilots in real time, with the aim of improving control, safety, human-machine interaction and decision support in highly automated aircraft. Small UAS develops an AI-driven swarm system made up of ten uncrewed aerial systems, aiming to improve endurance, autonomy, survivability and operational range.

Other projects focus on practical battlefield functions. IntSen2 uses AI to support imagery intelligence based on European space assets. AIDA, the Artificial Intelligence Deployable Agent project, develops AI-based cyber defence agents capable of autonomous and semi-autonomous actions across the cyber incident management lifecycle. SABER researches collaborative flying and ground robots for battlefield delivery and data gathering. LODESTAR II integrates augmented reality and AI into soldier systems, including a head-mounted AR display, a body-worn AI processor and an advanced AR weapon sight.

Together, these projects show that Europe is already integrating AI into surveillance, cyber defence, robotics, logistics, training and decision support.

The governance of military AI in Europe is complex. The EU AI Act, adopted in 2024, establishes a comprehensive risk-based framework for AI oversight in the internal market. However, it does not apply to AI systems used exclusively for military, defence or national security purposes. Defence and national security remain closely linked to Member State competence and public international law.

This exclusion is purpose-based. If an AI system is used only for military or defence purposes, it may fall outside the AI Act. If the same system is used for non-excluded purposes, such as civilian protection, border management or law enforcement support, the AI Act may apply. This matters because many AI systems are dual-use.

Even where the AI Act does not apply, other legal frameworks remain relevant. International humanitarian law governs the conduct of hostilities, including the use of new weapons and methods of warfare. AI systems that affect targeting or combat decisions should therefore undergo structured legal review before deployment and when their functions change materially. Export control rules may restrict the transfer of AI software, models, training data and technical know-how. Defence procurement law shapes testing obligations, cybersecurity requirements, audit rights and control over system updates. Data protection law may also apply where personal data is processed.

The European Defence Fund adds another layer. Projects involving autonomous weapons must ensure meaningful human control. In particular, systems without meaningful human control over selection and engagement decisions for strikes against individuals do not qualify for EDF funding. The EU is therefore willing to fund defence innovation but not fully autonomous weapons that remove humans from life-and-death targeting decisions.

NATO is also part of the picture. It adopted its first AI Strategy in 2021 and released a revised strategy in 2024, reflecting newer developments such as generative AI. EU-NATO cooperation increasingly covers emerging technologies, interoperability, cyber threats and hybrid threats.

The opportunities are significant. AI can improve situational awareness by processing satellite imagery, radar data, drone footage and other sensor information faster than human analysts alone. It can support training and simulation, help commanders test scenarios and improve logistics by predicting equipment failures before they occur. In cyber defence, AI can help detect anomalies and accelerate responses to attacks. Autonomous systems can support reconnaissance, battlefield delivery and data gathering while reducing risks for personnel.

AI also supports Europe’s push for technological sovereignty. Developing European AI capabilities can reduce dependence on foreign suppliers, strengthen the defence-industrial base and improve interoperability among European armed forces. The European Defence Fund reflects this priority by reserving part of its annual budget for emerging disruptive technologies, including AI.

However, the risks are equally serious. The most sensitive issue is whether machines should make life-and-death decisions. For many European policymakers, the answer is no. The European Parliament, Austria and other actors call for strict limits on lethal autonomous weapons systems operating without meaningful human control.

Legal accountability is another challenge. AI does not remove responsibility but makes it harder to trace. If an AI-supported system contributes to a wrongful outcome, responsibility must be assessed across the full chain: design, testing, procurement, deployment, supervision and operational use. This becomes more complex with adaptive systems whose behaviour may change over time.

There are also technical and security risks. AI systems can be hacked, spoofed, manipulated or fed false data. A misclassification in a military context can have far more serious consequences than an error in a commercial system. Hostile states and non-state actors may also use AI to enhance cyberattacks, drone operations, surveillance or disinformation.

Finally, Europe faces a resource challenge. Defence innovation spending remains far below that of the United States and funding is still fragmented across national systems. EU initiatives such as the European Defence Fund aim to pool resources and reduce duplication but keeping pace with larger global players remains difficult.

Europe is navigating a narrow path. It aims to innovate quickly enough to remain secure but carefully enough to preserve human control, legal responsibility and public trust. The debate is not simply about whether AI should be used in defence. It is about how, under what limits and with what safeguards.

The coming years will be decisive. AI may transform defence in ways comparable to earlier military revolutions. Public awareness and democratic debate will therefore be essential. Europe’s challenge is to embrace innovation where it strengthens security while ensuring that human judgement, legal accountability and core democratic principles remain intact.