This chapter provides an overview of the Dutch Ministry of Defence’s approach to artificial intelligence (AI) and data science, its strategic vision for 2035, and its efforts in developing, organizing, fielding, and operating defence AI. The Dutch Ministry of Defence (MoD) views AI and data science as crucial for the protection of the Dutch Kingdom and has set three ambitions for the future. These include technological advancement, information-driven operations, and becoming a reliable partner. The government commits to spend at least 2% of GDP on defence. The strategic knowledge- and innovation agenda 2021–2025 focuses on innovation in short cycles and gives direction to knowledge building, technology development and innovation (Ministerie van Defensie 2020c).

The Data Science and AI Strategy 2023–2027 highlights the Dutch MoD strategy for the next five years with objectives related to high-quality information technologies, data governance, personnel policy, and data-driven decision-making (Ministerie van Defensie 2023a). Regarding ethics and defence AI, an ELSA Lab (Ethical, Legal, Societal Aspects) is established to ensure responsible AI use within defence, focusing on ethical, legal, and societal aspects. In addition, the Netherlands hosted the REAIM 2023 summit on Responsible AI in the Military Domain (Government of the Netherlands Undated) and is an active contributor to NATO’s Data and AI Review Board (DARB) (NATO 2022), which aims to develop a user-friendly and responsible AI certification standard in the military domain.

AI is recognized as an essential component for the future of defence. Therefore, the Dutch MoD focuses on integrating these technologies into unmanned autonomous systems, military decision support, logistics, business operations, safety, and security. To develop these technologies, the facilitation of data-driven working, improving data management, investing in knowledge and expertise, and optimizing decision processes, have been identified as key issues that will be synchronized in an all-encompassing roadmap. Collaboration with international partners and maintaining meaningful human control over AI systems is emphasized and described as essential.

During fielding and operating defence AI, the Dutch MoD is prioritizing responsible use. Examples of fielded defence AI projects include robotic and autonomous systems within the Army, a DataLab, and utilizing advanced algorithms and machine learning (ML) techniques within the Royal Netherlands Marechaussee.

It is vital to train and educate personnel at all levels to understand AI’s impact, including legal and ethical aspects. The Data Science Centre of Excellence is a major player within the MoD, that aims to strengthen the knowledge base of the Netherlands Defence Academy (Ministry of Defence Undated) and to develop data science courses for the different educational programs within Defence.

This chapter demonstrates the Dutch MoD’s commitment to adopting AI responsibly while acknowledging its importance for national security and defence. It underscores the efforts in research, development, and collaboration with knowledge partners, universities, and the commercial sector to remain at the forefront of AI technology while upholding ethical and legal principles. All information in this chapter is based open-source data.

1 Thinking About Defence AI

The Dutch MoD considers AI and data science as two areas of development that will significantly impact the way the Dutch Kingdom is protected and defended. The Dutch MoD holds a human-centric view on AI as a capability multiplier. To gain new insights and support decision-making through AI and data science, the MoD invests in enhancing knowledge and collaboration with public-private partners, and in experimenting with AI and data science. Dutch thinking on defence AI is influenced and inspired by NATO in general and several partners in particular. The Netherlands echo the principles for responsible military use of AI, laid out in NATO’s first AI strategy.

1.1 Strategic Vision 2035

To ensure that the MoD can gain new insights and support decision-making through AI and data science in the future, the MoD identified three main ambitions for the following years (Ministerie van Defensie 2020b). These ambitions will positively influence the effectiveness and efficiency of the MoD and are based on an analysis focusing on possible future threats, such as cyberattacks, and problems, such as the availability of too much information, making it difficult or even impossible to filter, analyse and act on information.

First, the MoD is a technologically advanced organization that uses AI, data science, big data and (semi-)autonomous systems under meaningful human control. Such an organization requires personnel that focuses on innovation and is educated in (information) technology. This provides the opportunity to work more labour-extensive where possible and remain labour-intensive where necessary and desirable. The weapon systems and units of the MoD need to be modular and easily upgradable, so they can be integrated and combined with other compatible systems and units. It is expected that AI algorithms can support such force mix optimization considerations. When innovating, special attention is given to implementation as well as ethical and legal aspects. The Netherlands government has approved the national plan on the NATO Defence investment pledge, which commits the MoD to increase defence spending to achieve the NATO goal of 2% GDP.

Second, the MoD aims to attain and hold a prominent information advantage within both NATO and the European Union (EU). A focus is placed on information-centric warfare. A key concept in this regard is Information Manoeuvre (Reynolds 2020). It holds a prominent position in Dutch military discourse, playing an increasingly pivotal role in discussions. It centres on leveraging military information capabilities to shape the behaviour of audiences by creating impactful effects in the operational environment. As outlined in Dutch policy documents, a well-executed Information Manoeuvre is deemed crucial for establishing a military that is prepared for future conflicts and resilient over the long term (Ministerie van Defensie 2020a). This relates to the concept of Multi-Domain Operations (MDO), which recognises the need to synchronise Information Manoeuvre with other military operations to create effects. Both Information Manoeuvre and MDO are relevant in the context of the complex and rapidly changing operational environment (Townsend 2018). In addition, using data science and AI to analyse large volumes of data gathered by an increasing number of sensors, the MoD can effectively manage complex scenarios and operations and broaden and facilitate its deployment capabilities in increasingly complex multi-domain and predominantly hybrid contexts. Think of algorithms to support scenario analysis, image recognition in video data or algorithms for optimal route allocation to support data collection.

Third, the MoD is a reliable partner and protector. The Dutch MoD is more visible and enhances the provision of information to the parliament and public. This will improve awareness about security, so the Netherlands will stay alert to (hybrid) threats and build resilience against them. In addition, the national security architecture will become more robust.

1.2 Strategic Knowledge and Innovation Agenda 2021–2025

The strategic vision 2035 calls for strengthening the innovative capacity of Defence and collaboration within its knowledge and innovation partners. The strategic knowledge and innovation agenda (SKIA) pays special attention to the ability to innovate in short cycles and gives direction to knowledge building, technology development, and innovation. With SKIA, the Dutch MoD has a concrete plan for action regarding the development of key enabling technologies such as AI and robotics. It focusses on four areas (Ministerie van Defensie 2020b):

  • Strengthening the foundation of defence-specific knowledge;

  • Embedding innovation in the work environment, culture, and management;

  • Enhancing collaboration among knowledge and innovation partners of the MoD;

  • Strengthening collaboration within the Defence Knowledge and Innovation Chain.

To attain these goals, SKIA defines several AI specific research and technology domains. These constitute, among others, the use of data science with respect to detection and deception using signature management in the electronic warfare domain. Another domain considers sensor systems and their use in countering remotely piloted aircraft systems. In the command-and-control domain AI is being used for increasing the speed of data collection and analysis, among other uses.

1.3 Data Science and AI Strategy 2023–2027

Even though AI and data science are described in the strategic vision 2035 and SKIA 2021–2025, the MoD has developed a Data Science and AI strategy for 2023–2027, as an explicit foundation for the optimal use of AI and data science for the next five years (which will be updated every five years). In shaping its strategy, the MoD consulted both public and private partners, such as leaders in industry and transatlantic partners like the US. In addition, frameworks and guidelines are being developed to document the functioning of an algorithm, the choices made, and its use in the workplace. The MoD is closely involved in interdepartmental developments in this field, such as various research initiatives and policy instruments. There is also significant international attention on the responsible deployment of data science and AI in the military domain. The MoD aspires to take a leading role in international standards and certification developments. Optimizing ethical-legal frameworks and the responsible use of data science and AI in the military domain, in close collaboration with NATO and EU partners, research institutions like the Netherlands Organisation for Applied Scientific Research (TNO) and the Royal Netherlands Aerospace Centre (NLR), and industry, is a key focus for the MoD.

The Data Science and AI strategy considers five areas of application of AI that guide the development of AI together with strategic partners as NATO and EU (Ministerie van Defensie 2020b):

  • Unmanned autonomous systems;

  • Military decision support and intelligence;

  • Logistics and predictive maintenance;

  • Business operations;

  • Safety.

In its call to action the strategy identifies the need for further European cooperation on data science and AI. There are four common objectives that require an integrative approach regarding AI and data science:

  • Develop high-quality information technologies, to quickly and safely connect or share data within the MoD and with partners.

  • Uniformly organise data management and governance, so data can be securely and quickly shared within the MoD and with partners.

  • Develop personnel policy to invest in needed knowledge and expertise.

  • Work-processes and decision-making have a data-driven approach, by investing in high-quality information technologies that enable this.

1.4 Ethics and Defence AI

To focus on developing knowledge that is relevant for the responsible development and application of AI in the context of the entire lifecycle, ELSA-labs were created by the Netherlands AI Coalition (NL AIC) together with the Dutch Science Organization (NWO). The ELSA Lab Defence aims to develop an ecosystem for responsible AI use within defence and examines conditions under which AI applications are acceptable, emphasising ethical, legal, and societal aspects. It focuses on the use case of cognitive warfare and other non-traditional physical warfare. The ELSA Lab Defence is an ecosystem consisting of several knowledge institutes (universities) and (semi-)governmental defence partners. In addition, the ELSA Lab Defence coordinates with the other national ELSA Labs on findings and possible collaborations (such as the ELSA Lab Police) to establish a nation-wide narrative on AI ethics in the public-private sector.

The lab aims to develop an ecosystem for responsible AI use within defence. It proposes two main solution avenues: integrating ELSA factors into requirements, specifications, acquisition, and deployment processes (currently under active investigation), and educating military personnel, media, and policymakers about ELSA concepts. Existing approaches like “value-sensitive design,” “explainable” algorithms, and human-machine teaming are adapted for the defence context. Think for instance of the choice between complex machine learning models or simpler models that might be less accurate but better explainable in case of military combat. Realistic case studies such as terrain analysis for tactical helicopter missions are employed. The lab also studies public and defence personnel perceptions of military AI deployment, and tracks global technological, military, and societal trends influencing AI system deployment perceptions.

Simultaneously within the MoD, ethics, safety, and security must be integral components in the design process of data science and AI applications (privacy and security by design) to prevent vulnerabilities in the systems. It is crucial to consider data security and categorization, especially when linking systems for specific applications, as this introduces new challenges (Alfrink et al. 2022).

The MoD will therefore align its activities with similar initiatives related to secure data sharing brought forward by EU and NATO and beyond. In addition, the MoD opts for “privacy and security by design” and develops dynamic categorization for data science and AI applications based on combined data sources. The MoD continues to invest in knowledge building and technology development for new applications of data science and AI and invests in a robust analysis platform. A federative infrastructure for data science and AI will be developed. Furthermore, the MoD incorporates data science and AI as a focus area of cyber defence and develops a verification, validation, and accreditation system. The MoD also invests in knowledge building related to “counter-data science and AI.”

In 2023, the government of the Netherlands hosted the first global Summit on Responsible Artificial Intelligence in the Military Domain, REAIM 2023, together with South Korea. The REAIM 2023 summit was the world’s first global summit on responsible AI in the military and brought together governments, corporations, academia, start-ups, and civil societies to raise awareness, discuss issues, and to agree on common principles in deploying and using AI in conflict and war. Government representatives recognised not only the opportunities and potential but also the risks involved in using AI in the military domain. Therefore, they have agreed on a joint call to action on the responsible development, deployment, and use of AI in the military domain. The call-to-action focusses on the responsible use of AI in the military domain that follow international legal obligations in such a way that does not undermine international security, stability, and accountability. In 2024, the REAM conference is hosted by and organised in South Korea. It provides a good international platform in sharing knowledge and best practices about developing responsible AI in the military domain. In addition, it sets the international agenda on dealing with these technologies in the military.

Through its involvement in the REAIM summit, the Netherlands reinforced its dedication to ensuring responsible and ethical AI practices within the military that align with international law and respect of human rights. By leveraging its expertise in international law, the Netherlands contributes to shaping the global AI governance framework by emphasizing the importance of legal principles and maintaining accountability and transparency (under research in the ELSA Lab Defence among others) in the development and deployment of AI systems.

In 2023, NATO’s Data and Artificial Intelligence Review Board (DARB) has started the development of a user-friendly and responsible AI certification standard to help industries and institutions across the Alliance make sure that new AI and data projects are in line with international law, as well as NATO’s norms and values. The standard, which also applies to data exploitation and will include quality controls, is due to be completed by the end of 2024. Its aim is to translate NATO’s Principles of Responsible Use, approved in October 2021 as part of NATO’s first ever AI strategy, into concrete checks and balances, notably in terms of governability, traceability, and reliability. This will help to build trust among the innovation community, operational end users, and the public.

2 Developing Defence AI

The Dutch MoD envisions working in a “data-driven” manner by 2035. Data science and AI are critical enablers to conduct information driven operations in this regard. Several roadmaps, one for each of the five application areas discussed above, are being developed for this goal. It is imperative for the Netherlands to establish and maintain a strong position in the development and application of AI, especially with a strong focus within the MoD. By proactive deployment of AI-enhanced solutions it becomes possible to address the social and economic challenges that come with this new technology.

2.1 The Dutch AI Ecosystem

To realise and stimulate Dutch AI activities, the Dutch AI coalition (NL AIC) was established in October 2019 (NL AIC Undated). The NL AIC is a public-private partnership comprising government, business, educational and research institutions, and societal organizations, aiming to accelerate AI developments in the Netherlands and connect AI initiatives within the country. The ambition is to position the Netherlands at the forefront of AI knowledge and application for prosperity and well-being, while upholding Dutch and European norms and values. This is done by connecting traditional defence companies like Thales, IBM, and Deloitte for instance, in innovative ecosystems together with newer start- and scale-ups and universities. The traditional companies provide a solid basis upon which spin-offs and others can build and scale. The Dutch Defence Technological and Industrial Base (NLDTIB) consists of nearly 1000 companies with an annual turnover of €4.7bn. These companies focus on topics such as radar technology (Thales), sensor technology and/or international supply chains.

For the MoD this encompasses being at the forefront of each of the five identified application areas identified in the data strategy. Priorities are synchronized via a roadmap for each application area. The NL AIC acts as a catalyst for AI applications, with a key goal of achieving impactful AI innovations in at least ten economic and societal sectors within three years. AI is considered a systemic technology that requires an integrated approach involving intensive government involvement, urgency, proactive and comprehensive strategies, stakeholder engagement, and societal debate to determine how AI is deployed for prosperity and well-being.

On behalf of the Ministry of Education, Culture and Science, the Dutch Research Council funded research in the context of the Dutch Research Agenda (NWA) since 2018. This provides the Dutch MoD with the opportunity to invest in research focusing on data science with knowledge partners TNO, NLR and the Maritime Research Institute Netherlands (MARIN), and in knowledge building at universities such as both the technical universities (Delft, Twente, Eindhoven) and others with specific AI mission like Tilburg University’s focus on human-centric AI in understanding society. This builds a bridge between fundamental scientific research and applied knowledge building.

The MoD collaborates closely with the industry on various fronts to develop and apply data science and AI. This collaboration extends beyond the broader security domain to include other civilian application areas such as logistics, healthcare, and finance. Since 2021, the Dutch AI Coalition, in conjunction with Netherlands Industries for Defence and Security (NIDV) and TNO, has been providing additional opportunities for collaboration within ecosystems. Many start-ups and scale-ups are involved in a variety of research consortia and technical projects, varying from drone manufacturers to open-source predictive intelligence.

In addition, the Data Science Centre of Excellence (DSCE) was established, based on the cooperation of the Netherlands Defence Academy (NLDA) and the office of the Chief Information Officer (CIO) of the MoD. The aim is to enhance the scientific knowledge of the Academy in the field of data science and AI, by conducting research on data science and AI within the defence domain and developing data science courses for different educational programs. The research agenda will be based on topics that are fundamental for the future Defence organisation on the strategic, tactic and operational level, taking on a multidisciplinary perspective combining insights from the different research clusters of the Netherlands Defence Academy Faculty of Military Sciences. The research agenda will be the foundation for a long-term research and educational agenda. The DSCE will be located within MINDlabs—a participatory ecosystem of Tilburg University, several research labs, and companies—and closely cooperate both with partners within the MoD as well as outside (academic) partners (Ministerie van Defensie 2023b). Defence partners provide cooperation on long term research questions and opportunities for valorisation. Academic partners provide research collaborations as well as opportunities for shared education on data science.

2.2 Defence AI Development

As developments within AI and data science are occurring exponentially fast, the MoD intends to collaborate with EU and NATO partners, knowledge institutions and industry to keep up (Bharadiya 2023). The strategy contributes to a cohesive data and information field by connecting various developments within the organization through an overarching vision. As previously mentioned, the priority lies in developing AI for unmanned autonomous systems (UAS), military decision support and intelligence, logistics and predictive maintenance, business operations, safety, and security. For instance, unmanned autonomous systems are developed from a naval perspective with respect to maritime drones for intelligence, surveillance, and reconnaissance (ISR) missions, at the Army Robotics and Autonomous Systems unit with unmanned small ground vehicles combined with drone swarms, and at the Air Force to support manned-unmanned teaming. Military decision support with data science and AI is developed at all warfare centres (varying from Army, Air Force to Navy) and several data scientific units within organizational support units. Predictive maintenance applications for instance are being developed in collaboration with the knowledge centre smart maintenance and the Royal Netherlands Navy.

The MoD is investing in the development of a federated resilient IT infrastructure in which several networks can be coupled via satellite, wide area, or local networks to ensure connectivity and interoperability, and to ensure data sharing and processing. Simultaneously, it is implementing an ethical framework of norms for algorithms into internal policy. In addition, the DSCE is responsible for education on data science and AI for officers in training, and for academic research in an ecosystem of knowledge institutions and industry.

3 Organising Defence AI

In 2020, the Dutch MoD appointed a dedicated CIO, responsible for all data, IT, and cyber initiatives within the MoD. It recognized that the evolving nature of war, now encompassing digital battlegrounds, demands a focus on data science and cybersecurity. This includes safeguarding the nation and its allies against cyber threats. Traditional defence has expanded to include protecting against cyberattacks and misinformation campaigns, recognising the significance of digital expertise in national security. As state actors exploit virtual means to further their foreign policy goals, adapting to this changing landscape is crucial, and the ability to gather and utilise accurate and timely information is deemed essential for success. The CIO partakes in the Executive Board of the MoD, highlighting the importance of technology expertise in strategic decision-making. The MoD is also prioritising education and training in cybersecurity and data science for its staff, recognising the role of data-centric warfare in shaping future conflicts.

To efficiently organise the goal of the MoD to become a data-driven organization in 2035, several foundational steps are being taken. First, data-driven working will be facilitated across all units of the MoD but implemented in a decentralised fashion. The MoD aims for a model in which units are supported with their data-driven work by use of Defence-wide facilities, provided amongst others by the IT command, in the following three areas: (1) technology and tools to support the process from data analysis to production; (2) specialised knowledge and expertise; and (3) ethical frameworks, guidelines, and clear process steps (governance). To remain innovative, it is important that data initiatives are carried out bottom-up within the respective Defence units and coordinated across the services by the decentralised CIOs of the respective service. Involving the workforce and utilising domain-specific knowledge ensures that initiatives align with requirements of the specific units and by doing this contributes to Defence-wide data knowledge.

The second organisational step is to have all data management appropriately organized. In addition to technological capabilities, well-defined data management is a crucial prerequisite for achieving the ambitions of the MoD. Improving data quality and ensuring “one single version of the truth” are crucial for optimal data utilisation and the implementation of reliable data products (Dragt 2020). This is achieved in a federated structure where the central CIO provides the foundation upon which the decentralised CIOs of the services build using one generic platform. Access to data sources with associated concepts and definitions must be uniformly regulated and provided by the central CIO and coordinated by the decentralised service level CIOs. There will be agreements on data and algorithm ownership and responsibilities, as well as rules and conditions for using (aggregated) data sources in a Defence-wide or decentralised analytical environment. These agreements are risk-based (complex datasets using complex algorithms being highest risk). The quality of data largely determines the quality of data products and is therefore an essential part of data management.

The third organisational step is to invest in knowledge and expertise. To keep advancing in the realm of new technology and data, it is important that Defence invests broadly in knowledge and expertise (Hartmann and Henkel 2020). This is done by utilising existing knowledge and experience optimally. For instance, researchers at the NLDA that conduct research in the domain of data science and AI are connected through the newly created DSCE tasked with research but also education to increase the awareness and knowledge on this topic within the MoD. Different Defence units are at different maturity levels with respect to data science and AI. The lessons learned and experiences of those that are more mature can be used as growth nuclei to professionalise and scale through knowledge sharing. The DSCE is set up to become the main hub on education and research on data science and AI for the MoD.

The final adjustment to the organisation of the MoD is to evaluate and optimise decision processes and protocols. To use modern data technology, work and decision-making processes must be scrutinised if a significant increase in relevant data for decision-making is expected. This involves the creation of hybrid man-machine teams and integrating data-based decision-making into existing or new protocols.

It is necessary to look beyond the impact of adopting new data and technology. Business and military operations must become aligned with data-driven work. This requires the adaptation of existing doctrines which can only be done step by step, by continuously improving military training, and changing business operations and workflows based on lessons learned on specific cases. One approach in this regard is to involve researchers from the DSCE as advisors in the doctrine formulation process and associated projects. This is done by active collaboration between researchers from the DSCE and the specific warfare centre of the relevant operational commands.

4 Funding Defence AI

Unclassified details regarding budget and spending of the MoD show that since 2020, Defence has established a material budget fund of approximately €66bn for the next 15 years. This enables the MoD to better plan its expenses, including the management and maintenance of investments such as infrastructure. As a result, Defence is investing in a high-quality, future-proof military and defence organisation.

In the previous years, the Netherlands was one of the countries of NATO that had the lowest spending of its MoD in research and technology development. However, for 2023, the Dutch government has increased its defence budget by 40% relative to the 2022 budget and approved an additional €5bn in budget for the MoD on a structural basis (Ministry of Defence 2022). This will mean that the Netherlands will meet the NATO’s 2% of GDP standard by 2024 and 2025. However, investments in data science and AI are still limited compared to those of the US and China (Roberts et al. 2021).

Various measures described in the Defence Industry Strategy (DIS) will be intensified in the following years by €252M. The DIS describes the base of what is needed in the interest of national security and is another initiative to improve the technological striking power of the MoD and help transforming the armed forces. With an increased budget, the MoD will, for example, further invest in Research & Technology and in so-called short-cycle innovation, thereby strengthening ecosystems relevant to the MoD. In addition, the Netherlands will invest €56M into the NATO Innovation Fund that is being created for technology development using start-ups and scale-ups.

Specifically focusing on AI, a long-term program called AiNed, set up by the Netherlands AI Coalition, aims to strengthen the AI-position of the Netherlands. The program focuses on large-scale projects that tackle bottlenecks to enable AI opportunities. The National Growth Fund of the Dutch government has allocated €204.5M to the AiNed program 2021–2027. The National Growth Fund aims to allocate a total of €20bn for the period 2021–2025 to projects that focus on knowledge development and research, development, and innovation, since these two fields have the highest potential for economic growth (Rijksoverheid Undated).

5 Fielding and Operating Defence AI

The Dutch MoD has been actively prototyping, fielding, and operating defence applications of AI. AI technologies are very promising, but the MoD is cautious in fielding them. Awareness of the dangers is important, and AI should be used responsibly. The MoD wants to act in line with international humanitarian law, and closely follows international developments around AI legislation. In addition, data requirements and data processing aspects are explicitly considered in the procurement processes of future weapon platforms and systems to ensure access to data and compliance with standards. Certification is considered on a per technology basis, depending on procurement requirements or other guidelines adopted for internally developed technology.

There is a limited amount of Defence AI projects being fielded and used that can be shared. As the second international partner to receive the F-35, the MoD is increasing its knowledge on the use of AI in modern weapon systems by actively engaging with the supplier. Most defence AI projects are still being researched and/or under development. We describe three distinct examples of defence AI that are currently actively being fielded and operated by the MoD below.

5.1 DataLab

In June 2021, the Datalab was officially launched within the IT-support command of the MoD. The Data Lab is intrinsically motivated to contribute to the peace and security of the society daily. Everything DataLab is therefore aimed at, is making the Dutch MoD a global leader in the performance of its main tasks. Since DataLab sees a crucial role for data science and AI in a highly diverse range of problems, the lab plays a leading role across the MoD to unlock that potential. To achieve this, the work of DataLab consists of four main processes:

  • With its innovation process, DataLab identifies the challenges of tomorrow and the solutions for them in a structured manner.

  • DataLab operationalises promising innovations in such a way that their contribution to the main tasks of Defence is made tangible.

  • Datalab enables the Defence-wide development and production of applications by making tooling, standards and a secure and managed LGI platform (Defence Wide Data Development and Analysis Platform, DBDAAP) available.

  • As a network organiser, DataLab facilitates and encourages a knowledge sharing and connection process across the Defence organisation to accelerate the development of meaningful applications.

Against this background, DataLab works on enterprise defence AI and AI for military use as the following two projects illustrate:

  • MLOps (enterprise defence AI)

The MoD experiences a surge in machine learning (ML) demand amidst limited resources. Addressing this challenge hinges on efficiency. One way to achieve this is by adopting the MLOps process, which ensures that Data Science operations are transparent, reproducible, and meet stringent quality standards. Proper governance of the respective Data Science environment is essential, and these factors are key in achieving it. Through the introduction of a dedicated MLOps tool, the DataLab provides both low-code and high-code environments to the Data Science community. This tool empowers a greater number of users to engage in Data Science and participate effectively in the MLOps process.

  • Submarine detection with multispectral satellite images (AI for the military use of defence)

This project focuses on investigating the possibilities of developing an improved detection method for submarines (as well as other objects and troop build-ups) using multispectral satellite images. For this purpose, a multispectral dataset is combined with ML image recognition algorithms.

5.2 The Robotics and Autonomous Systems Unit

The Dutch MoD has a Concept Development and Experimentation program on Robotic and Autonomous Systems (RAS) with the purpose of identifying the best combination of organisation, concept of operations, and unmanned systems to increase the combat power of land units and increase the protection of soldiers. Its unit is a subunit of the Dutch Army, with the following key topics: combat unmanned ground systems, swarming unmanned aerial systems, and autonomy.

Autonomy in unmanned systems is defined in various ways, for instance as the ability to execute ordered tasks within the set constraints and conditions and with delegated levels of decision-making authority (Antsaklis 2020). Automating functions such as navigation decreases the cognitive and physical burden of the human operators. Autonomy in unmanned systems is necessary to ensure the continuation of task execution when the datalink between the operator and the unmanned system is jammed or unavailable. Autonomy also facilitates upscaling to many unmanned systems performing ordered tasks that are supervised by a few humans. Maintaining meaningful human control over unmanned systems with autonomy is an essential element of the Dutch Army policy.

The RAS program started with the first practical experiments with the Milrem THeMIS and Rheinmetall Mission Master Unmanned Ground Vehicles (UGVs) in 2019 and deployed an infantry platoon with military qualified combat unmanned ground vehicles to Lithuania as part of NATO’s enhanced forward presence mission to conduct further experimentation, validate the concept of operations and identify the process for operationalisation of innovation. The RAS program combines applied research, conceptual thinking, and practical experimentation in close connection with the operational 13th Light Brigade. Their unmanned systems provide the opportunity to increase the combat power of military units. The development and integration of AI functions increases the capabilities of the unmanned systems.

The two main intelligent functions are autonomous movement and aided target recognition. The development of autonomous movement includes global and tactical route planning, local path planning, obstacle detection and avoidance, and dynamic route re-planning on the UGV. Aided target recognition includes object detection and classification by computer vision algorithms, target tracking and reporting, and target environment assessment. The development of autonomy also encompasses tactical planning by an automated intelligent planner, plan visualization and course of action comparison and monitoring and dynamic plan adjustment during execution.

5.3 Deep Vision

By utilising advanced algorithms and ML techniques, the Royal Netherlands Marechaussee (RNLM) can process information more rapidly and accurately, enabling proactive measures against criminal activities. These criminal activities range from cross-border human smuggling or attacks on V.I.P.’s, both in the Netherlands and abroad. This strengthens the capabilities to secure the Kingdom of the Netherlands.

The Deep Vision team of the RNLM experiments with new technology in the field of sensing, robotics, large language models and other types of AI, and 3D-printing. The aim is to gain knowledge, to explore how innovative technology can facilitate the daily operations of the RNLM, and to provide insight into long-term implications of emerging technologies for the RNLM.

Developing in-house AI capabilities guarantees an independent position of the RNLM (and the MoD in general) in relation to commercial tech-suppliers and other partners. This ensures flexible and interoperable AI-deployment within the RNLM. In addition, given the possible ethical complications around the deployment of AI, it is essential for the RNLM to thoroughly understand both the benefits and the pitfalls of new technology. Therefore, it is crucial for the RNLM to not solely concentrate on the advantages of emerging technologies but also proactively anticipate potential adverse consequences. Close collaboration with other (inter)national Defence components and partners within the security domain (such as the Dutch Police) is essential for the exchange of technical knowledge and skills to further develop AI-capabilities within this domain. The RNLM also collaborates with several high-tech universities, research institutes and tech companies to experiment with the most advanced AI and data science technology of the present day.

6 Training for Defence AI

The digital transformation and increasing datafication have a great impact on the work of military personnel and civilians (Mattila 2022). Therefore, one of the main points of attention of the MoD is having high-quality AI knowledge and skills. (New) training courses focusing on technology, the effects for military deployment, and associated legal and ethical dilemmas increase knowledge and insights among employees at all levels of the MoD.

For example, one expert group affiliated with the MoD develops simulation-based training using AI as a player in the simulation is the Royal Netherlands Aerospace Centre (Royal NLR). Royal NLR develops training and simulation programs for military forces and other clients. Their goal is to make training more effective and efficient by using integration and interoperability of live, virtual, and constructive elements. One example of such a training are tactical simulators for fighter pilots in which AI is used to increase the intelligence of the behaviour of opposing forces (OPFOR).

TNO’s Defence, Safety, and Security unit runs several labs across the country and is focusing on innovation for defence and national security with an emphasis on combating crime, catastrophes, and terrorism. The involvement of this unit covers a range of activities: military operations, military equipment, command & control, and operational decision making, threat protection, and instruction and training. Among other TNO Defence focuses on enhancing the effectiveness and efficiency of armed forces by studying human factors, optimizing training through simulations, and experimenting with new security concepts and equipment. They also specialize in improving human-machine collaboration.

The unit focusses on four areas (TNO Undated):

  • Operations and human factors, by supporting the Dutch armed forces with innovative analysis and evaluation methods, models, and the latest simulation technologies. Involved in the ELSAlab defence.

  • Information and security systems, by devising and developing technology for an effective management of the military organisation and the deployment of resources.

  • National safety, by focusing on five main themes: Resilient professionals, Smart security & surveillance, Intelligence in action, Rule of law & investigation, and Critical digital infrastructure.

  • Protection, munitions, and weapons, by offering innovations for civil and military forces, from effective strategy through material innovation to weapon and system testing.

In the pursuit of secure and dependable Data Science and AI applications, the MoD collaborates closely with the private sector as mentioned before. One such example is their partnership with the Dutch AI Coalition (NL AIC), a “public-private collaboration involving governmental bodies, businesses, educational and research institutions, as well as civil society organizations, aimed at expediting and fostering AI advancements and initiatives” (NL AIC Undated).

The MoD also develops knowledge and skills from within by working together with the Defence Academy (NLDA), thereby strengthening the link between scientific research and defence practice. The NLDA is unique in the Netherlands as it combines (maritime) military education, university-level education, and personal development. Knowledge and skills in the field of data science and AI are being integrated in military training courses of the NLDA and in continuing training both for civilian and military employees of the MoD. Some examples of training focusing on data and AI at the NLDA includes the masterclass Cyber & Data Science and the “Data” course, which educates participants on data & ethics, data & military decisions, and data & intelligence. In addition, there is a “Data” module at the Open Defence Academy. This online module is about data and its importance for defence, where concepts such as algorithms, data science, and AI are discussed.

Furthermore, different research, education and training initiatives aim at increasing knowledge and skills regarding data science and AI within the MoD. Representative examples include the Data Science in Military Operations Chair at the Faculty of Military Sciences of the MoD in conjunction with the endowed chair in Data Science, Safety, and Security at the Tilburg University in the Netherlands, and the DSCE, which is headed by Chair of the Data Science in Military Operations group (Ministerie van Defensie 2023b).

These chairs contribute to the advancement of data science in the military and security domain. Research of the chair is conducted against the backdrop of a changing world with significant implications of the use of AI and digitization. It focuses on ethical algorithm development and aspects of the use of AI in the security domain, algorithms to support (military) decision-making, and intelligence and cyber security.

The DSCE develops data science courses for different educational programs of NLDA. The research agenda will centre around essential themes crucial for the future of the Defence organisation, spanning strategic, tactical, and operational levels. This approach adopts a multidisciplinary perspective, integrating insights from various research clusters within the Netherlands Defence Academy Faculty of Military Sciences.

7 Conclusion

This chapter provides an overview of the Dutch MoD’s approach to AI by highlighting the strategic vision for 2035, the strategic knowledge- and innovation calendar 2021–2025, the Data Science and AI strategy for 2022–2027, and the commitment to ethical considerations in the development and deployment of defence AI. The Dutch MoD recognizes the pivotal role of AI and data science in safeguarding the Dutch Kingdom and has set three key ambitions: technological advancement, information-centric operations, and becoming a reliable partner and protector. These ambitions align with the commitment to invest at least 2% of expenses of the Defence budget in research and technology development.

Data-driven decision-making and responsible AI use are important for the Dutch MoD, as demonstrated through the creation of ELSA Labs and the hosting of the REAIM 2023 summit on responsible AI in the Military Domain. In terms of developing defence AI, the Dutch MoD is actively engaged in collaborating with EU and NATO partners, knowledge institutions, and industry across various domains, including unmanned autonomous systems, military decision support, logistics, business operations, and security. It is also investing in IT infrastructure, ethical frameworks, and education to ensure responsible AI development and deployment.

The organization of defence AI within the MoD includes steps to facilitate data-driven work, improve data management, invest in knowledge and expertise, and optimise decision processes. The report highlights the practical fielding and operation of defence AI, with a focus on responsible use and adherence to international humanitarian law. It provides insights into specific projects, such as the DataLab, the Robotic and Autonomous Systems, and showcases the Deep Vision team’s efforts to utilize AI and data science for national security.

Training for defence AI is a critical aspect, with various educational initiatives aimed at equipping personnel at all levels with the necessary knowledge and skills to navigate the evolving landscape of AI and data science. Overall, this report underscores the Dutch MoD’s dedication to adopting AI responsibly while prioritizing national security and defence. It demonstrates a commitment to research, development, collaboration, and ethical principles, positioning the Netherlands as a leader in the responsible use of AI in the military domain.

Some of the challenges that lie ahead emerge from the fact that AI technologies, public sector innovation, and legal and regulatory developments do not evolve in tandem. Innovative techniques to mitigate biases inherent in AI systems are being developed. The increasing intensity of working with machines requires optimal human-machine teaming, interfaces, and processes. Finally, strategic alignment across (inter)national partners is required to guarantee long-term sustainability of current and upcoming AI systems, as they need continuous updates, maintenance, and improvement. Investments into these systems should optimally align with defence strategies and priorities.