Photo. CyberDefence24/Canva
Artificial intelligence is no longer only an experimental support tool. It is moving into military command, intelligence, logistics and autonomous systems. Its real value will depend not only on fast algorithms, but also on trusted data, interoperability, resilient communication and keeping human responsibility for decisions.
Artificial intelligence is becoming an important component of military transformation because modern operations generate more information than a staff can analyse in the available time. Multi-domain operations require coordination of land, air, maritime, space and cyber activities, while the information environment affects all of them. Advantage is increasingly connected not with one platform, but with the ability to combine sensor data, identify useful patterns and translate them into coordinated action. In this context, AI is gradually entering command-and-control, intelligence processing, logistics, cyber defence, targeting support and management of autonomous platforms.
One of the main contributions of AI is shortening the distance between detection and decision. Data from unmanned aircraft, satellites, radars, electronic intelligence, open sources and cyber networks can be combined into a more complete operational picture. Machine-learning systems filter repetitive information, identify anomalies and propose links which may not be immediately visible to an analyst. This does not remove human judgement. It changes the conditions in which a commander makes a judgement, because prioritised information and several courses of action can be produced much faster than in traditional staff work.
NATO’s revised AI strategy and the Alliance Digital Strategy adopted in 2026 reflect this change. Data governance, secure sharing, artificial intelligence, modelling and simulation are now treated as connected foundations of digital transformation. Synthetic environments are expected to support collective training, planning, wargaming and mission rehearsal. The task is not simply to buy more AI applications. NATO needs a federated ecosystem in which Allied systems can exchange trusted data and use algorithms without losing security, interoperability and political control.
For Central and Eastern European countries, this has a very practical meaning. The eastern flank is exposed to short warning time, electronic warfare, cyberattacks and disruption of space-based navigation or communication. AI-supported systems can improve recognition of preparations for aggression, management of dispersed forces and situational awareness. Their effectiveness will still depend on data quality and resilient networks. A sophisticated algorithm connected to unreliable communication does not give decision superiority and can create another dependency.
Planning, decision and support
Operational planning is one of the fields where AI may bring benefits relatively quickly. Algorithms can compare many variables, such as location and readiness of forces, ammunition, weather, terrain, intelligence indicators and possible reactions of the opponent. This makes it possible to generate several variants and update them when the situation changes. The commander still defines the objective and accepts the risk. The system supports assessment of how the available resources can be used, but it should not replace operational judgement.
The U.S. Army Project Convergence shows how this model is being tested in practice. It connects sensors, command networks and effectors across services and with multinational partners. Capstone experiments examined AI-assisted situational awareness, data-driven decisions and reduction of the sensor-to-shooter time in operational scenarios. The announcement of Capstone 6 in 2026 showed that the programme is moving to larger experimentation, and not only separate technology demonstrations. Its main importance is the effort to check whether many different systems can work as one operational architecture.
A comparable direction is visible within NATO’s Defence Innovation Accelerator for the North Atlantic. In 2026, DIANA launched a Decision Superiority for NATO Warfighters challenge designed to identify AI, machine-learning and software tools for modelling, simulation, targeting support and operational wargaming. The selected solutions are intended to demonstrate how fragmented data can be transformed into usable knowledge for Allied Command Operations. This is particularly important because NATO operates as a multinational alliance in which information is produced by different national systems, under different security rules and in different technical formats.
The problem, however, is interoperability. A national AI system may work well and still have limited value in coalition operations if it cannot receive or understand Allied data. Common metadata, interfaces, testing rules and access control are necessary, but political agreements on classification and information sharing are also needed. During a crisis, the speed of decision will depend on willingness to share data almost as much as on computing power. Technology cannot solve institutional barriers by itself.
AI can support predictive analysis, although this term should be used carefully. Military behaviour does not follow stable rules in the same way as physical processes. An opponent can deceive sensors, change tactics and exploit assumptions built into a model. Predictive tools should indicate possibilities and warning signs, rather than promise a certain forecast. Their usefulness is greater when they are combined with red-teaming, human analysis and constant verification against operational evidence.
Autonomy, logistics, cyber defence and the space domain
The connection of AI with autonomous and uncrewed platforms is one of the most visible directions of military innovation. Collaborative combat aircraft are intended to operate with crewed platforms, extend sensor coverage, conduct electronic warfare and provide additional mass in high-risk missions. The MQ-28 Ghost Bat is an example of this approach. Recent trials showed more complex cooperation with command-and-control and combat aircraft, including autonomous mission functions and weapon employment. Such systems can reduce risk to personnel, but they need reliable communication, clear command relations and protection against deception or hostile takeover.
AI-supported autonomy also matters for land, maritime and logistics operations. Uncrewed vehicles can transport supplies, conduct reconnaissance and work in contaminated or heavily contested areas. In logistics, algorithms may forecast demand, identify possible equipment failures and optimise routes according to weather, terrain and enemy activity. This can improve availability of critical systems. Nevertheless, wartime logistics cannot be organised only according to efficiency models. Redundancy, reserves and the ability to work without digital support remain necessary, because the most efficient network can also become the easiest to disrupt.
In cyber defence, AI is used to identify abnormal behaviour, prioritise alerts and automate selected responses. Military and civilian networks generate such a large number of events that manual analysis alone is insufficient. The same technology, however, can support offensive activity by automating reconnaissance, generating convincing phishing material, identifying software vulnerabilities or adapting malicious code. The competition is therefore symmetrical: AI strengthens defence while lowering the cost and increasing the scale of attack. The protection of training data, models and software supply chains becomes part of military security.
The space domain creates similar opportunities and risks. AI can process imagery and signals on board satellites, reduce the volume of data transmitted to ground stations and detect changes in maritime or land activity. It can also support collision avoidance, satellite tasking and the management of constellations. This is relevant to multi-domain operations because space services provide communications, navigation, missile warning and intelligence to forces in every other domain. A loss of access to these services may have immediate consequences for operations on land, at sea and in the air. Resilience therefore requires distributed architectures, alternative navigation systems and the ability to continue operations when individual satellites or ground stations are unavailable.
Human control, legal responsibility and trusted military AI
The expansion of military AI creates legal and ethical questions which cannot be solved only by technical certification. The most difficult questions concern systems able to identify, prioritise or engage targets with limited human involvement. International discussions on lethal autonomous weapon systems continued in 2026, but there is still no universally accepted legal instrument. International humanitarian law applies, although uncertainty remains about the required degree of human control, responsibility for mistakes and the ability to explain decisions produced by complex models.
NATO adopted six Principles of Responsible Use: lawfulness, responsibility and accountability, explainability and traceability, reliability, governability and bias mitigation. Their implementation needs more than a political declaration. Systems should be tested in degraded communication, during adversarial manipulation and with data different from the training environment. Commanders have to know the limits of the system and when its recommendations may be unreliable. Governability has key importance, because a system may work correctly in technical terms and still produce an operational effect which is politically unacceptable.
Data quality is another strategic vulnerability. Biased, incomplete or deliberately corrupted data can lead to incorrect classification and flawed recommendations. This problem is especially serious in coalition operations, where data originate from different sensors and national procedures. A trusted AI ecosystem therefore requires secure data provenance, common evaluation methods and continuous monitoring after deployment. Certification cannot be treated as a one-time event because both the operational environment and the software may change.
Doctrine, training and organisational adaptation
The use of AI will also change the organisation within the armed forces. Command structures need personnel who understand not only operational requirements but also the limits of data-driven systems. This creates a need for mixed teams of commanders, intelligence specialists, software engineers, legal advisers and testing experts. Military education should prepare officers to question algorithmic recommendations, recognise uncertainty and compare machine-generated options with experience. Without such competence, commanders can either reject useful systems or trust their outputs too easily.
Procurement procedures will also need to change. Traditional acquisition programmes are designed for platforms with relatively stable configurations, while AI-enabled software develops through frequent updates and continuous learning. Armed forces must be able to test new versions, control access to models and retain the ability to operate when a commercial supplier is unavailable. Open architectures can reduce dependence on a single company, but openness must be balanced with cybersecurity and the protection of intellectual property. The state should avoid a situation in which it owns the platform but lacks effective control over the data, software or maintenance process that determines its operational value.
Training should include deliberate failure of AI-supported systems. Units need to practise with incomplete data, corrupted models, interrupted cloud access and contradictory recommendations. This is especially important on the eastern flank, where electronic warfare and attacks against communication infrastructure may appear already in the first stage of a crisis. A resilient force should use AI when it works, but keep analogue procedures, human expertise and delegated command when digital support is degraded.
Artificial intelligence will not replace command, strategy and political responsibility. More probably, it will redistribute tasks between humans and machines. Systems will perform filtering, correlation, optimisation and some autonomous functions, while people will define objectives, interpret ambiguity and remain responsible for the use of force. The greatest advantage may not belong to forces with the biggest number of algorithms. It will belong to those which connect technology with doctrine, training, reliable communication and clear rules of control.
For NATO, the transformation of multi-domain operations should proceed on technological and institutional levels at the same time. Common data standards, interoperable architectures, experimentation and faster adoption are one part. Legal clarity, political supervision, human competence and preservation of command responsibility are the second part. AI can shorten decision cycles and improve coordination, but it also creates new points of failure. Maximum automation should not be the strategic objective. More important is a reliable decision advantage in difficult and disrupted operational conditions.
