The Dutch government has developed a vision because of the widespread availability of generative AI and the scale and pace of its current development. While the technology offers many opportunities, the government also sees risks. Irresponsible use or development can jeopardize public values and fundamental rights such as transparency, privacy, autonomy, and non-discrimination. The Netherlands has adjusted its vision on generative AI in the short term. In 2023, the government distanced itself from commercial generative models like ChatGPT and DALL-E, preferring open-source versions of the technology. This vision document outlines how the use of generative AI is permitted (regardless of form) and encouraged across the entire government spectrum.

Dealing with the opportunities and challenges of generative AI is primarily a task that must be addressed jointly, intergovernmental and intersectoral, with a learning and experimental approach. This means continuously conducting a broad
societal dialogue in the Netherlands and seeking international cooperation based on this, within the EU and globally.

The government-wide vision on generative AI emphasizes the importance of governance and a value-driven approach within the broader digitalization policy. Four principles are applied:

1. Generative AI is developed and applied in a safe manner.
   
   Actively contributing to mitigating misuse, accidents, and systemic security risks of and by generative AI models.
   
   
2. Generative AI is developed and applied in a fair manner.
   
   This includes considering risks of inequality of opportunity, discrimination, transparency and explainability, and contributing to promoting equality and bridging (socio-economic) divides.
   
   
3. Generative AI must serve human well-being and safeguard human autonomy
4. Generative AI contributes to sustainability and our prosperity
   
   Generative AI is used to promote sustainable economic growth, reduce labour shortages and lead to innovative new products and solutions, including solutions to societal issues such as climate change.
   
   
   

The above principles are in line with the Sustainable Development Goals. The Netherlands aims to play a leading role in Europe and the world by creating conditions for responsible development and use of generative AI while maintaining our digital open strategic autonomy.

The approach to translating the vision into action includes six pillars:

1. Collaboration
   
   

   Generative AI poses fundamental challenges that require close collaboration and societal dialogue at various levels. Involving residents, implementers, employees, employers, entrepreneurs, businesses, and professionals in shaping the future of generative AI is essential.

   Concrete action points:
   
   
   1. Solidifying and deepening existing international collaborations.
   2. Potentially joining the Alliance for Language Technologies European Digital Infrastructure Consortium (ALT-EDIC).
   3. Appointing an organization to stimulate and coordinate initiatives around societal dialogue on generative AI.
   4. Promoting participation in the development of generative AI models by and within the government.
   5. Exploring and experimenting with the implementation of responsible generative AI applications in administrative processes and government services.
   6. Researching the use of generative AI to improve government information management.
      
      
2. Actively following developments
   
   The government aims to prepare for the expected exponential growth of generative AI by training civil servants and closely monitoring relevant developments. Special attention is given to the implications of generative AI for the labour market, with the Socio-Economic Council tasked with mapping out its impact on labour productivity, quantity, and quality of work. Additionally, the impact of generative AI on sustainability and climate is being investigated, acknowledging that while this technology can contribute to addressing climate challenges, it currently leads to an increase in the ecological footprint.
   
   Concrete action points:
   
   
   1. Establishment of a new advisory body (Rapid Response Team AI) to advise the Ministry of Digital Affairs and the State Secretary on important developments.
   2. Establishment of a transparency register for initiatives within the government involving generative AI.
   3. Research on how generative AI can contribute to citizen participation and communication with citizens.
      
      
3. Shaping and applying legislation and regulations
   
   The government will commence the broad implementation of the EU AI Act within the public sector in 2024 and will focus on providing guidance and education to stakeholders.
   
   Concrete action points:
   
   Development of a Dutch regulatory sandbox
   
   
4. Increasing knowledge and skills
   
   The government actively seeks to increase knowledge and skills within the public sector to fully harness the potential of generative AI. This includes providing guidelines for the deployment of generative AI within government organizations and tightening procurement conditions. A knowledge centre for generative AI within the government can coordinate these developments, while investments in AI talent and technological infrastructure ensure the preservation of European norms and values. Additionally, the importance of media literacy and critical thinking in education is emphasized, supported by research into the impact of generative AI by the National Education Lab AI (NOLAI).
   
   Concrete other action points:
   
   A training program is being developed for civil servants, with special emphasis on ethical issues.
   
   
5. Innovating with generative AI
   
   The Netherlands must foster an environment that allows for testing, scaling, and improving reliable and transparent (generative) AI models and tools, including research into establishing a national AI testing facility. The importance of reliable Dutch-language datasets as the foundation for generative AI models is increasing, while a National AI Validation Team facilitates publicly available benchmarking and tooling for the responsible use of generative AI in the Netherlands. Concurrently with the development of the open language model GPT-NL, efforts are being made to develop ethical guidelines for the responsible use of generative AI.
   
   Concrete action points:
   
   Establishment of AINed Innovation Labs (public-private labs focused on the development of AI innovations).
   
   
6. Strong and clear supervision and enforcement
   
   Finally, the government emphasizes the importance of strong and clear oversight and enforcement. A proactive approach, good cooperation between regulators and other stakeholders, and continuous evaluation of policies and laws are essential to prevent and manage undesirable effects of generative AI in a rapidly changing technological landscape.