*Last updated: August 2026. This page explains why we believe our work helps in preparing the Netherlands for transformative AI. It is deliberately a causal argument, not a programme list. For more information on what we do and where, see our chapters.*
Our mission is to develop the talent the Netherlands needs to address the risks of transformative AI. This page explains how we aim to accomplish this and what would have to be true for us to be wrong.
The situation
We believe that on the current trajectory AI systems that can do most economically and strategically important cognitive work, including AI research itself, may arrive within three years, and very likely within ten. Importantly, whether that happens or goes well is not decided yet. It depends on how many capable people work on making it go well, and right now far too few people are working on it. The critical question is therefore not only what AI will become, but whether enough people are ready to shape what happens next.
The Netherlands is unusually well placed to matter here. We have world-class universities, ASML at the centre of the global semiconductor supply chain, a strong tech sector, proximity to Brussels, emerging organisations working on AI safety and security, and a government that has an international AI strategy and needs more capacity to execute it. The opportunity is to turn these existing strengths into a pipeline of people who can do meaningful AI safety work, and build that capacity here in the Netherlands.
The complication
The Dutch problem is not a lack of potential talent. It is the lack of a clear path from interest to capability to meaningful contribution. This pushes capable and motivated people to either relocate to AI safety organisations in London or San Francisco, or they drift back into other careers (sometimes ones that make the problem worse). Meanwhile demand for AI safety talent in Europe is rising sharply, and the institutions that need it are being defined in the next few years. People who want to contribute need somewhere to start, a community to learn with, and a path to doing serious work. At the same time, European institutions are building their AI safety capacity now, but the people needed to staff them are not yet available at sufficient scale.
We see four gaps in the Netherlands: local group infrastructure, skill, awareness, and institutional support. Awareness is rising, and we expect it to continue doing so. The more important gap is what happens next: a path from concern to capability into action. We also believe the institutional gap is dependent on the infrastructure and skill gaps. Institutions can only build effective AI safety functions when capable talent exists. That leaves two gaps that will not close without deliberate work: local group infrastructure (people and local AI safety groups being disconnected from each other, and everyone reinventing everything from scratch) and skill (very few people in the Netherlands do real AI safety work). These two are our focus.
Our bet
The causal chain is simple to state:
- If we build a strong national talent pipeline with local SAIN chapters supported by shared infrastructure like a common playbook, courses, events and a project hub, then capable people get a clear path from first hearing about AI safety to making their first real contribution.
- If we then connect those people to the wider field through warm referrals, fellowship and application support, and connections to institutions and organisations, then they end up in positions of leverage: the potential future NL AI Security Institute, ministries, EU bodies, AI safety organisations, or new organisations they may create themselves.
- If enough people take this path, the Netherlands develops the talent and institutional capacity necessary to be prepared for transformative AI, and can help lead the group of middle powers pushing for safe development and a pause or pacing of unsafe development internationally.
We focus on the first half of this pipeline, from no awareness (level 0) to first real contribution (level 3), and the handoff after it. Events and outreach take people from 0 to 1. Courses and discussion groups take them from 1 to 2, on both a technical and a governance track. The project hub takes them from 2 to 3. The national layer then connects level-3 people to the destinations above us.
We do not pretend to own the full path, and we work hard not to lose people at the handoff. Progressing to the highest levels of the pipeline often depends on the opportunities available in the wider AI safety field: universities, governments, research institutions, AI labs, and other organisations must open roles for safety talent to apply and deepen their skills.
What we count
Our primary measure of success is how many people make a first real AI safety contribution through SAIN each year, and how many we successfully hand off toward full-time roles or new organisations. A first real contribution is a verifiable output, e.g. completing a Project Hub project, co-authoring a publication or policy output, winning a prize in a hackathon, or facilitating a course.
We will also track progression through the pipeline: how many people move from awareness to education, from education to contribution, and from contribution into further work in the field.
What has to be true
Every theory of change rests on assumptions. Ours are these four:
- Demand materializes. The NL AI Security Institute, EU AI Act operationalization and the Dutch International AI Strategy create real roles in the coming years.
- The talent is here. Dutch universities and industry produce enough capable people, provided a path exists.
- Conversion works. Our courses and Project Hub reliably turn interest into first contributions. We will be adamant about testing different conversion mechanisms.
- Handoffs stick. People we refer to where there is demand actually land in the field and stay there.
We will track all four and ensure that if anything changes or breaks we change the plan. We are always looking to receive feedback and are open to change course if the evidence suggests the need to do so.
How we are organised
SAIN is a registered Dutch foundation (stichting) with local chapters, currently in Groningen, Amsterdam and Utrecht. The foundation provides the shared infrastructure the chapters run on. This will happen similarly for future SAIN chapters.
For SAIN's mission, pillars, and long-term ambition, see the Vision.