GPT-NL: A Sovereign Language Model for the Netherlands

The Netherlands Organisation for Applied Scientific Research (TNO), SURF, and the Netherlands Forensic Institute (NFI) are collaborating to build GPT-NL, an independent language model optimized for the Dutch language and local context. The project aims to reduce technological dependence on specific countries or companies and strengthen digital autonomy.

The core values of GPT-NL are as follows:

• Sovereignty: Directly controlling the model and data within the Netherlands and Europe, moving away from dependence on non-European providers.
• Transparency: Documenting the data collection and training process, releasing the source code as Open Source, and providing the model weights under a controlled license.
• Trustworthiness: Training the model from scratch to avoid the data issues of existing models, while strictly adhering to personal data anonymization and copyright protection.
• Reciprocity: Aiming for a fair innovation ecosystem through a model that shares revenue with data providers.

This project has received a total of 13.5 million euros (approximately 20 billion KRW) in public funding from the Netherlands Enterprise Agency (RVO), under the Ministry of Economic Affairs and Climate Policy.