Cabinet sets out AI plan, seeks partners beyond US and China

The Dutch cabinet’s AI vision is a plan for fitting artificial intelligence into society while capturing its benefits and limiting strategic dependence on foreign powers. It also treats AI as a national-security issue. Digital economy minister Willemijn Aerdts said AI will “largely determine our future,” making choices about leadership and limits urgent.

The plan imagines AI giving nurses more time with patients, helping entrepreneurs grow and supporting responses to major challenges. It does not require the Netherlands to do everything alone. Instead, it seeks the ability to choose its own direction and work with other EU countries and selected international partners.

The vision names Canada, Japan, Singapore and South Korea as partners. It also supports stable relations with the United States and China. The cabinet has joined Pax Silica, a US-led partnership, but its wider plans still need funding. Only €120 million is identified so far, while infrastructure, talent and research require more investment.

“Middle powers” generally means countries with meaningful economic, technological or diplomatic influence, but without the global dominance associated with superpowers. The term is not a formal ranking. In this plan, it describes countries that can cooperate internationally while giving the Netherlands more options beyond the United States and China.

Canada, Japan, Singapore and South Korea are named as possible partners, alongside other EU member states. Working with them could support shared approaches to AI, technology and security. It could also help the Netherlands avoid putting all its strategic dependence on one or two larger powers.

The cabinet still wants “stable and predictable relations” with the US and China. It has also joined Pax Silica, a US-led partnership covering chips, computing power and AI infrastructure. The vision therefore combines established relationships with a wider network of technologically capable partners. The article does not specify separate projects or agreements with the four middle powers.

So far, the cabinet has identified €120 million for Dutch participation in an EU-wide AI project. That money was already set aside in September. It is the only funding figure included in the cabinet’s 28-page vision letter, so the document does not present a complete financial programme for Dutch AI policy.

The letter says further “targeted and structural investment” is necessary. It specifically points to infrastructure, talent and research. These areas cover the practical capacity to develop and run AI, the people who can build and govern it, and the knowledge needed for future systems. The article gives no amounts for those additional needs.

That gap has drawn criticism. Volt leader Laurens Dassen called the plan “a toothless tiger” and said AI cannot be a top priority without a budget. More detailed plans for computing power and datacentres are expected in the first quarter of 2027. Their costs and funding are not yet stated.

If the Netherlands relies too heavily on the United States or China, decisions made abroad could affect its access to chips, computing power and AI systems. That could weaken Dutch control over technology, raise exposure to supply disruptions and make national security more vulnerable. These are reasonable strategic risks behind the cabinet’s stated aim of controlling reliance on foreign powers.

The mechanism is dependence: if essential hardware, infrastructure or systems come mainly from another country, that country can influence availability, conditions and direction. The article specifically notes Dutch participation in Pax Silica, a US-led partnership on chips, computing power and AI infrastructure. It also says the Netherlands will maintain stable, predictable relations with China.

The cabinet is seeking balance rather than isolation. It wants cooperation with Canada, Japan, Singapore, South Korea and EU members, while developing Dutch infrastructure, talent and research. However, the vision provides only €120 million so far. Plans for computing power and datacentres are still due in the first quarter of 2027.

Chips, computing power and datacentres form the physical foundation of AI. Chips carry out the enormous number of calculations involved in training models and producing results. Computing power means the available processing capacity. Datacentres bring together the servers, storage, cooling and network connections needed to operate these systems.

During development, AI systems process large datasets repeatedly to learn patterns, which requires powerful chips and substantial computing capacity. During use, the same infrastructure turns a request into an output such as text, a prediction or a recommendation. More capacity generally allows more users, larger systems and faster responses. Datacentres provide the controlled environment where this equipment operates.

The article connects this foundation to national strategy. The Netherlands has joined Pax Silica, a US-led partnership covering chips, computing power and AI infrastructure. It also says plans for Dutch computing power and datacentres will follow in the first quarter of 2027. Those facilities consume large amounts of electricity and water, creating practical and environmental constraints.

A Dutch AI safety institute could provide specialised expertise on how AI affects security. It might test systems, identify dangerous capabilities and advise the government, although the article does not yet define its exact powers. The proposed model is the UK’s AI Security Institute, and an advisory report is due in December.

The media pact would address a different risk: how AI systems use journalistic content. Aerdts and the education minister plan to agree it with public and private media organisations. Such an agreement could create common expectations for handling news material, but its mechanism and rules have not been announced. The article gives no details on issues such as licensing or access.

Both proposals are still under development. The safety institute is only being explored, pending the advisory report. The media pact is also a planned agreement rather than an operating policy. Together, they show the cabinet considering institutional responses to AI risks in security and journalism, alongside its broader national AI vision.

Artificial intelligence is a broad term for computer systems that perform tasks commonly associated with human intelligence, such as recognising patterns, generating language or making predictions. Many modern systems learn from examples rather than following only hand-written instructions. This general definition is established knowledge; the article discusses AI’s uses but does not define the term.

A learning system is given data and an objective. During training, it adjusts internal parameters to capture useful relationships in those examples. When new input arrives, the system applies those learned patterns. It can then estimate what may happen, produce likely text, rank options or recommend something. The output is generated from the system’s learned calculations, not human understanding in the ordinary sense.

This explains why the cabinet links AI to both opportunity and risk. AI could give nurses more time, help entrepreneurs grow and address major challenges. At the same time, powerful systems require chips, computing power and data, and can affect national security and journalism. The Dutch vision aims to guide those effects.