Artificial Intelligence (AI) is a collective term for technologies that, based on mathematical models, replicate human capabilities. ‘AI’ covers techniques such as pattern recognition, generative AI and large language models (LLMs). AI is changing our society and government work at every level. Its consequences cannot yet be fully foreseen. AI will touch every implementation process, so it needs an integrated policy to manage it.
The government wants to harness AI’s potential responsibly to address societal challenges and improve services for citizens and businesses. At the same time, it aims to ensure responsible deployment by regulating transparency and safety, upholding democratic oversight, and maintaining service quality amid a tight labour market.
Central government, provinces, municipalities, water authorities, and public service providers are leading with a single guiding approach, making successful AI applications usable across government and removing barriers to innovation so scaling up becomes possible. They are building an AI infrastructure independent of non-European AI developments. Business and science also need to be deeply involved. Partners working together on the NDS make responsible AI use possible across government.
The objectives
The objectives of the NDS AI priority are twofold. First, government organisations harness the opportunities of AI in a responsible and decisive way to improve their operations and services for the citizens and entrepreneurs of the Netherlands. Second, the Dutch government has a high-quality AI infrastructure, consisting of (among other things) high-quality data, open language models from the Netherlands and/or the EU, and sufficient computing power. The infrastructure also supports talent training and retention.
The accelerators
The NDS’s AI priority has identified 6 accelerators:
3.1 AI Scaling-up Facility
The AI Scaling-up Facility is an approach to increase the impact of deploying and using promising or proven AI applications in government. It identifies barriers and offers support to make scaling up promising AI possible. In this way, the government is working towards a single guiding approach for scaling up AI applications. It uses AI opportunities responsibly to improve public service processes and the services provided to citizens and entrepreneurs.
A guide, Scaling up AI, has been published, containing a step-by-step plan to help governments weigh decisions on scaling up AI.
3.2 Priority AI application areas
Selected application areas where work is done jointly to develop at least one AI application per area, for possible scaling up within the Scaling-up Facility’s approach. The priority application areas focus on using AI in government. They cover AI for ‘objections and appeals’ (with a view to simplifying legislation), ‘AI and the civil servant’ (aimed at increasing civil servants’ productivity, such as personal AI assistants), ‘AI and the Open Government Act’ (including AI tools that save time in WOO processes) and ‘AI and service delivery’ (improving services for citizens and entrepreneurs).
3.3 Auditable Algorithm Framework & AI procurement approach
To deploy AI and algorithms responsibly and comply with legislation, clear and uniform frameworks are needed across government. Joint auditable standards for government use of AI are therefore being established in an Auditable Algorithm Framework. Experts are currently validating the auditable framework. In addition, the AI procurement approach sets out supplementary procurement conditions and guidelines for the purchase of AI systems by government, so that they comply with the AI Act and align with existing procurement conditions such as ARBIT, and so that government organisations procure AI systems faster, more carefully and more responsibly.
3.4 Government-wide AI Competence Centre
The lack of specialist AI knowledge and expertise within government is a bottleneck to the responsible use of AI. This exploratory study aims to identify which expertise is needed, already exists and must be shared, and to define the form an AI competence centre should take. The study aims to map how government can bring together, expand and unlock AI expertise, and, in due course, share capacity as well.
In practical terms, this means working towards a better-organised, government-wide AI expertise base, so that knowledge, support and capacity are deployed more effectively, and so that government can work in a more challenge-oriented way in the field of AI.
3.5 AI infrastructure and building blocks
A public platform infrastructure offers governments a central place to develop, configure and share building blocks and applications for generative AI for reuse, in a secure and responsible manner. This delivers efficiency (develop once, use many times). It also strengthens government’s digital autonomy and its strategic collaboration with business.
A government-wide marketplace has been launched (working title: sAIdkick), where government organisations can find, share and reuse existing AI solutions and knowledge. The marketplace offers AI applications, reusable building blocks and knowledge products, supported by a rapidly growing community of 450+ professionals and an AI broker who brings supply and demand together. At the same time, government is strengthening its AI infrastructure, among other things by connecting to existing facilities including the AI factory in Groningen.
3.6 Encouraging (open) NL/EU language models
Encouraging the use of open or European closed language models is essential. It will improve transparency, reliability, ethical accountability, and digital independence. We must avoid creating a risky strategic dependency on AI. This requires a new government policy on the use of generative AI models. In other words, we are expanding the deployment and use of language models that better align with our values. Work is underway on a vision for using generative AI models in government. Completion is scheduled for the end of 2026.
