Appendix. Scenarios for the Next Three to Five Years: Strategic Foresight Methodology | From Plausible Tomorrows to Prompt Action on AI

How are today’s and tomorrow’s developments in AI going to have an impact on the Netherlands? The honest answer is: nobody knows – including (future) AI itself. Why? Because the future is fundamentally unknowable. This is true in a general sense and even more true when it comes to developments that are taking place at a vast pace and that are having a disruptive impact in nearly all thinkable domains. AI is such a rapidly evolving, high-impact field, affecting the political, economic, military and societal realms. This makes forecasting impossible as well as detrimental. To allow for sound policymaking despite that, we may apply foresighting.

Strategic foresight offers us the possibility to evaluate not one but several perspectives of possible or Plausible Tomorrows that could reasonably materialise. In other words, it is not about what will happen (forecasting) but what could happen. This study offers three scenarios that can help stretch our thinking in terms of what could be on the AI horizon and to offer perspectives on what that could mean for national security. If we want a Dutch AI and national security strategy and policy that are future proof, we need a strategy and policy that are futures proof.

The three Plausible Tomorrows are derived from expert discussions in a scenario workshop hosted at the Clingendael Institute in May 2025, with participants from think tanks, the Dutch government and the AI sector. The aim was to imagine the national security implications of four different scenarios across two axes of fundamental uncertainties. The four resulting ‘quadrants’, as we called them, were supplemented with a third dimension: the degree of availability of open-source AI models. The three scenarios that differ most from today’s situation are the three Plausible Tomorrows presented in this report.

Two Uncertainties, Four Quadrants

Participants in the workshop were presented with a Dutch version of the following visual representation of the resulting quadrants:

The four quadrants were summarised as follows:

Quadrant 1: The Netherlands misses the ‘AI boat’ and lies by the wayside, while others prosper. In this conceivable future, developments within and around AI have been less disruptive over the past few years. The Netherlands, like other EU Member States, has (very) limited access to AI and the very latest technologies arising from it. Open-source AI has developed to a limited extent. Closed-source AI systems, especially from US and Chinese companies, dominate the market. Of the four conceivable futures in this study, the one described in quadrant 1 most closely resembles today’s status quo.

Quadrant 2: Because of open source, everyone goes ‘before the wind’: to what extent is the Dutch ship really seaworthy for such a scenario? In this conceivable future, open-source models have continued to develop, albeit at a limited pace by AI standards. Open source has thus become dominant. This has made AI, as well as important AI applications, more widely accessible than ever before: not only state actors but also non-state actors have substantial capabilities and can thus develop a wide variety of AI applications, whether for political, social, economic or military purposes. In short, the Netherlands has the capabilities, as does (potentially) everyone else.

Quadrant 3: ‘The Netherlands is sailing, but in very turbulent waters’; (rapid/disruptive development of AI and more access of AI). In this conceivable future, developments within and around AI have developed very disruptively over the past few years. This is the case with respect to closed AI, as well as open-source models. In addition, tremendous progress has been made in the development of agentic AI, allowing this technology to be used in all kinds of applications and hardware (such as smart devices). As a result, AI has come to play a very comprehensive role on the political, social, economic and military stage. The US and China occupy a dominant position where closed military AI systems are concerned. Europe, including the Netherlands, the Gulf region and countries such as Brazil and Japan also have closed AI models that (can) be competitive. At the same time, companies and governments are struggling to keep their heads above water in this highly competitive and rapidly changing world. Widely available open-source models and agentic AI are adding to this and are having an impact in several sectors, such as education, healthcare, policing and for many knowledge workers.

Quadrant 4: ‘The Netherlands is at the mercy of the AI (weather) gods’. In this conceivable future, developments within and around AI have developed very disruptively over the past few years. AI has come to play a very comprehensive role on the political, social, economic and military stage. Only the US and China have highly advanced AI. Other countries, including the Netherlands, are entirely dependent on one (or both) of these (AI) superpowers in this regard. Widely available open-source models are inferior to the closed models of the US and China in terms of capabilities. Agentic AI is vastly developed in this scenario, with far-reaching consequences, such as the widespread deployment of humanoid robots.