Artificial Intelligence (AI) is progressing rapidly and increasingly affecting government sectors. While the technology offers opportunities for better service delivery and more efficient processes, it also poses risks, particularly regarding security, autonomy, and public values. As part of the Netherlands’ Digitalisation Strategy (NDS), AI remains one of the 6 key priorities.
Under the leadership of Larissa Zegveld, the acceleration team is developing a joint approach to enable all levels of government to use this technology responsibly.
Zegveld, who serves as Director General of the Kennisnet Foundation and Chair of the Netherlands Standardisation Forum, has been involved with the NDS since its inception. When it became clear that the strategy would continue despite political changes, she committed to her role as chair. “I saw a strong collective will to tackle this together. The time is now. This will only succeed if we are willing to collaborate differently,” she explains.
The importance of investing in AI
Zegveld views AI as more than just a technological innovation; it is a strategic factor intertwined with broader geopolitical and economic developments. “Our dependence on foreign suppliers is growing rapidly, raising the question of how the Netherlands and Europe can maintain their autonomy. We talk about sovereignty, but if we mean it, we must act now. The opportunity to develop alternatives will only come once,” she warns.
She remains optimistic about the Netherlands’ capacity to develop these alternatives, highlighting the increasing number of experiments and initiatives. She also observes a rising willingness within Europe to collectively invest in technology that aligns with European values. “The Netherlands is a leading country in digital standards. We have repeatedly shown that we can influence suppliers to adhere to our norms. This gives me confidence that we can carve out our own position.”
Framework for responsible use
AI can bring substantial benefits to the government, from quicker service delivery to smarter support for professionals. However, innovation requires strong frameworks. Government organisations manage personal data and deliver socially essential services, making careful design and use of AI crucial. “We deliver public services that users cannot opt out of. This comes with extra responsibility,” Zegveld emphasises.
The AI acceleration team is therefore focusing on shared conditions. European regulations, such as the AI Act, offer important foundations, but these need to be adapted to the Dutch context and implemented in procurement, deployment, and management. “We need to clarify what we mean by responsible deployment. This requires standardisation and practical translation,” she adds.
The governance challenge
For Zegveld, the biggest hurdle is not technical but governance-related. Breaking down institutional silos is essential to achieving collective solutions. “This is not a technical problem; it’s a governance challenge. We only achieve scale in alternatives like Vlam.ai if we stand behind them together. This requires different behaviour and the courage to cross boundaries.”
The Netherlands has a strong tradition of independent implementing organisations and departments, each with its own responsibilities. This can make collaboration more complicated than in countries with more centralised governance. Nevertheless, Zegveld sees change taking place. The NDS priorities clarify where the focus should be in the coming years. “We have identified these six topics as important. That also means saying no to things that don’t fit within this framework.”
The AI Acceleration Team: Speeding up and connecting
Within the acceleration team, CIOs and directors from implementing organisations, ministries, municipalities, provinces, and water authorities collaborate. This ensures they are involved from the outset; a deliberate choice, according to Zegveld. “The solutions must land there, so they need to be part of the design and realisation. They know better than anyone what is feasible.”
The team’s approach emphasises tangible progress. 6 accelerators have been established, outlining what needs to be done and the desired outcomes. “We’ve stopped perfecting documents on paper,” Zegveld says. “The plans are good enough to get started. That in itself is a change. Now, we’re actually going to do it.”
In the upcoming period, teams will be assembled to deliver results for each accelerator. The acceleration team will also review ongoing initiatives that can be embraced or expanded, including Vlam.ai, GPT-NL, and the recently launched AI Factory. The aim is to identify which solutions align with shared frameworks and where the government can act as a ‘launching customer’. According to Zegveld, “This is how we ensure alternatives truly mature.”
Strategic objectives for AI in government
The AI priority has 2 strategic goals:
-
Improving public services and tackling societal challenges with AI: Zegveld emphasises the importance of implementing AI carefully. The government manages data that citizens are required to provide and bears responsibility for delivering reliable service. “This means we must approach new technology differently than commercial firms. Innovation is permitted, but it must be responsible.”
-
Building a high-quality AI infrastructure aligned with public values: This involves developing trustworthy models, ensuring access to high-quality datasets, securing sufficient computing power, and training and retaining talent. A collective approach is essential. “We’ve made it clear that non-commitment is no longer an option. We are moving toward shared solutions that everyone uses. Individual solutions won’t get us further.”
Connection with other priorities
AI does not operate in isolation; it is directly and intensively connected to cloud, data, and cybersecurity. AI models only work effectively when the correct data is available in sufficient quality. At the same time, it must be clear where this data is stored, how it is secured, and whether it remains within the appropriate legal framework. “For years, we’ve been working on single data collection and multiple uses. The same applies to AI. You need to know exactly where the data is and where it’s going,” Zegveld explains.
The dependence on cloud services directly ties to the Cloud priority. A sovereign AI stance is impossible without managing the underlying infrastructure. Consequently, leaders of the acceleration teams collaborate closely. “We meet in the NDS Council and the Programme Management Consultation. This helps prevent silos and promotes coherence.”
Looking ahead, and a message to the reader
Within a year, Zegveld anticipates that scaled-up applications will be widely adopted. This means others will adopt successful solutions from one organisation, and shared provisions will become the standard. “Scaling up is key. That’s where we need to go.” Her central message is clear: the time for talk has passed. “We’re going to make it happen. Non-commitment is no longer an option. If we use big words like autonomy and sovereignty, we must act accordingly. This can only be achieved if we change the way we collaborate.”
