Dutch Media Companies Collaborate on Their Own AI Language Model, GPT-NL
den haag, vrijdag, 5 september 2025.
More than 30 Dutch media companies are working together with TNO on GPT-NL, a Dutch variant of ChatGPT. The aim is to reduce the power of large tech companies and promote reliable Dutch language skills. The media companies are making their archives available to improve the AI model and create a transparent alternative that complies with European legislation. Herman Wolswinkel of NDP Nieuwsmedia emphasises that news articles not only add language but also provide reliable world knowledge through verified facts.
Collaboration for a Transparent Alternative
This collaboration between more than 30 Dutch media companies and TNO is a response to the growing dominance of large tech companies in the AI sector. By sharing their archives from the past 20 years, researchers from TNO and partners can further develop GPT-NL. The goal is to create a transparent and fair alternative to ChatGPT that complies with European legislation and counters the dominance of big tech [1].
Investment in Reliable World Knowledge
Herman Wolswinkel, director of NDP Nieuwsmedia, stresses that news articles not only add language to AI models but also provide reliable world knowledge through verified facts. ‘News companies can contribute to innovation without compromising their news media,’ says Wolswinkel. Participating companies invest 400 million euros annually in editorial teams to deliver reliable news [1].
Impact on the Media World
The development of GPT-NL has both advantages and potential drawbacks. In the long term, participating media companies expect a language model that does not infringe on copyright and supports the editorial process. However, in the short term, they hope to gain little. According to Steef de Bruijn, editor-in-chief of the Reformatorisch Dagblad, they do not expect participation to yield significant financial gains [1].
Ethical Considerations and Safety
Despite the benefits, researchers are also focusing on the ethical aspects and safety of AI technology. Lian Tian and Marian-Andre Rizoiu of the University of Technology Sydney discovered vulnerabilities in AI models that allow for the creation of disinformation through simple prompts. They advocate for stricter security measures and transparency from AI companies [2].
Future Applications and Training
The development of GPT-NL is also a step towards digital sovereignty in the Netherlands. Saskia Lensink of TNO emphasises that the project aims to develop, strengthen, and sustain digital sovereignty. Upcoming workshops and events, such as the Data Expo 2025 in Utrecht, offer opportunities for professional development and knowledge sharing on the latest advancements in data and AI [3][4].