Dutch news publishers contribute to developing GPT-NL

Dutch news publishers are collaborating with the Netherlands Institute of Applied Scientific Research (TNO) to develop further GPT-NL, the first large-scale Dutch AI language model trained on legally obtained data. The partnership between NDP Nieuwsmedia members and TNO aims to create an AI model that respects copyright and public values, setting a precedent for ethical and responsible AI development that does not exploit the work of journalists.

GPT-NL is an open-source language model designed for the Netherlands, emphasizing transparency, inclusivity, and adherence to Dutch and European values. The project, which began in 2023, is a collaboration between TNO, the Netherlands Forensic Institute (NFI), and SURF, a supercomputer company. This initiative aims to provide a reliable and secure tool that benefits Dutch society, particularly in sectors that handle large volumes of sensitive documents. GPT-NL is designed to understand the nuances of the Dutch language, culture, and legal system, ensuring data protection and digital sovereignty.

Ethical AI development

The collaboration with NDP Nieuwsmedia, representing the majority of Dutch news publishers, marks a significant step in the development of GPT-NL. Members of NDP Nieuwsmedia are providing access to their archives, which include articles from over 30 national and regional news titles. This contribution is expected to double the amount of high-quality Dutch data used for training the model. News agency ANP is also participating, making this the first instance worldwide where private news publishers are collectively collaborating with an organization to develop an AI model. This partnership underscores a commitment to responsible AI development, respecting copyright and public values.

The initiative directly addresses concerns about Big Tech companies training AI models on news articles without permission or compensation. By partnering with NDP Nieuwsmedia, TNO is demonstrating an alternative approach that promotes ethical AI innovation while strengthening journalism in the Netherlands. According to NDP Nieuwsmedia, this collaboration sets a precedent for advancing AI responsibly, without infringing on the rights of journalists. The GPT-NL project also prioritizes data protection, excluding confidential and harmful content, and avoiding duplication, ensuring a fair and transparent dataset.

GPT-NL in the making: a language model based on Dutch law

With GPT-NL, the Netherlands has its own language model starting this year.

Read More

Looking ahead

Training for GPT-NL began in June 2025, with plans to refine the model for initial use in the fourth quarter of 2025. With access to over 20 billion tokens from NDP Nieuwsmedia, GPT-NL is poised to become a valuable tool for various applications, particularly within the security domain. The project not only aims to enhance the digital sovereignty of the Netherlands but also contributes to an autonomous Europe in 2040, where AI operates in accordance with European standards and values. Earlier contributors of data to GPT-NL include De Nederlandsche Bank (DNB), ICTRecht, and Het Utrechts Archief.

Rethinking intelligence: Is AI making us dumber?

As AI reshapes our world, Zednik suggests intelligence isn't waning—it's evolving. AI is challenging us to redefine and thrive within our ever-changing technological landscape.

Read More